E-BOOK

GaN Power Amplifier

Design Solutions

SEPTEMBER 2018

S P O N S O R E D B Y
Table of Contents

3

Introduction
Pat Hindle
Microwave Journal, Editor

4 Space Fence Radar Leverages Power of GaN

Justin Gallagher, Joseph A. Haimerl, Thomas Higgins and Matthew Gruber
Lockheed Martin MST, Moorestown, N.J.

10

Solid-State PAs Battle TWTAs for ECM Systems
Rick Montgomery and Patrick Courtney
Qorvo, Greensboro, N.C.

14 GaN Powers Microwave Point-to-Point Radios

Kristoffer Andersson, David Gustafsson and Jonas Hansryd
Ericsson, Gothenburg, Sweden

19

Mastering the Thermal Challenges of Advanced
Defense Subsystems
Duncan Bosworth and Gary Wenger
Analog Devices Inc., Norwood, Mass.

23 Going Green: High Efficiency GaN Amplifiers

Patrick Hindle, Microwave Journal Editor
Microwave Journal Editor

29

A Wideband High Efficiency Doherty Power Amplifier Based
on Coupled Line Architecture
Guangping Xie , Zongxi Tang, Biao Zhang and Xin Cao
University of Electronic Science and Technology of China

2
Introduction

GaN Power Amplifier Design Solutions

According to Strategy Analytics, the RF GaN market growth continued to accelerate in
2017, with revenue growing at over 38 percent year-on-year. With GaN seeing adoption across a
variety of RF applications, the rollout of commercial wireless infrastructure coupled with demand
from military radar, electronic warfare and communications applications are providing the primary
drivers for growth. They forecast that RF GaN revenues will exceed $1 billion by the end of 2022,
with defense sector demand slightly greater than commercial revenue.
Strategy Analytics reported that RF GaN demand from the military sector grew by 72
percent year-on-year in 2017, and they project it will grow at a compound annual average growth
rate of 22 percent through 2022. The military radar segment remains the largest user of GaN
devices for the defense sector, with substantial production activity in AESA radars for land-based
and naval systems being the main driver for increasing demand of RF GaN.
They also note that wireless base stations continue to be the largest revenue segment
for RF GaN, with increasing market penetration resulting in year-on-year growth of more than
20 percent. While the big lift from Chinese LTE deployments is over, the wireless industry has
maintained and, in some cases, compressed the 5G deployment schedule. The resulting 5G base
station deployment will become a primary commercial growth driver for RF GaN.
With RF GaN amplifier activity continuing to increase, this eBook provides six articles
covering various design topics in commercial and military markets. The first two stories cover
Aerospace & Defense applications with GaN being used in the Space Fence Radar System and as
replacements for TWTAs in ECM systems. The next article takes a look at the adoption of GaN in
the point to point microwave radio systems authored by Ericsson. With the high power densities
achieved by GaN amplifiers, the thermal challenges can even be more critical than the electronic
design issues so the next three articles look at addressing thermal challenges through heat
sinking/mechanical design and various design techniques and solution to improve efficiency. We
hope that this will improve your knowledge of GaN amplifier technology and thank Boonton and
Qorvo for sponsoring this eBook.

Pat Hindle, Microwave Journal Editor

3
Space Fence Radar Leverages
Power of GaN
Justin Gallagher, Joseph A. Haimerl, Thomas Higgins and Matthew Gruber
Lockheed Martin MST, Moorestown, N.J.

Editor’s Note: Because of its high power density, GaN is widely recognized as providing a step-function
increase in the capability of solid-state power amplifiers. Arguably not as well known is the impact
GaN is having at the system level, particularly military systems. While logical to assume GaN enables
evolutionary improvements, in some cases the system would not be possible without GaN. The Space
Fence radar is one example, described in this article by Lockheed Martin.

G
aN is a compound semiconductor
using III/V group elements that rep-
resents a leap ahead of existing GaAs
technologies for many RF applica-
tions. GaN high electron mobility transistors
(HEMT) are constructed using very similar fabri-
cation techniques as GaAs RF devices but pro-
vide significant advantages.
The reason for the leap in technology GaN
has over GaAs results from significantly higher
operating RF power density and higher reliabil-
ity. The secondary benefits built on these first
two are: smaller chips for a given output power;
smaller modules; higher overall efficiency due to
lower combining losses; lower module assembly
cost; higher power handling survivability and
higher impedance for large devices, making for
easier impedance matching.
The inherent advantage of GaN over GaAs
is due to basic physics of the devices and ma-
terials. GaN is referred to as a wide bandgap
semiconductor because of the wider energy gap
between the valence and conduction bands,
compared to conventional semiconductors. The
wider energy gap (3.4 eV for GaN compared
to 1.42 eV for GaAs) allows higher electric field
strengths and higher breakdown fields, which
translate into higher operating voltages.1
Another advantageous characteristic of GaN
is electron drift velocity, which in GaN increas-
es with higher field strength, while it decreases
with higher field strength in GaAs. Another way
of describing it is that the low field drift velocity
of GaAs is superior to GaN, but the high field
drift velocity of GaN is far superior to GaAs. The
peak in this characteristic (electron drift veloci-
ty vs. field strength) is the saturation velocity. In
GaN it is 2.46 × 107 cm/s, while GaAs is 1.8 ×
107. However, at high fields where GaN is 2.46
× 107, GaAs electron drift velocity is under 1 ×
107.1 Since electron drift velocity is related to
current density, it basically means that at high
voltage, GaN is also capable of high current,
whereas GaAs is not. Also, since power is a func-
tion of voltage and current, the wide energy gap
and high drift velocity make for ideal high power
devices. For these reasons, GaAs will never out-
perform GaN for high power applications; how-
ever GaAs should retain its applicability in low
voltage and low power applications.
A third important aspect of GaN technology,
especially for power devices, is the substrate

WWW.MWJOURNAL.COM/ARTICLES/26872
4

1E13
1E12
1E11
GaAs, TQPHT25
GaN, TQGaN25
Median Life (h)
1E10
1E09
1E08
1E07
1E06
1E05
1E04
1E03
100 125 150 175 200 225
Channel Temperature (°C)
s Fig. 1 Reliability comparison of Qorvo GaN and GaAs PHEMT
processes.
s Fig. 2 Space Fence comprises two radars, one located near the
equator at Kwajalein Atoll and an optional second site in Western
Australia.
material. Since GaN substrates were not very advanced and
were thermally inferior, silicon carbide (SiC) emerged as the
best substrate material for GaN RF power devices. The
thermal conductivity of SiC is superior to other substrates,
such as silicon or sapphire, and allows the high power po-
tential of GaN to be realized. The thermal conductivity of
SiC is approximately 10× better than GaAs. These advan-
tages allow operation of systems and sensors that would
not have been possible with GaAs semiconductors. For ex-
ample, new systems have steep requirements for both per-
formance and operational availability. GaN enables sensors
to operate with longer reliability and be driven harder (i.e.,
longer pulse width per duty cycle), based on the inherent
higher power density properties.
Assessing reliability, GaN devices can operate at a much
higher channel temperature than GaAs for equivalent reli-
ability; at equivalent channel temperatures, GaN achieves
much higher reliability. Comparing the reliability of Qorvo’s
GaN and GaAs technologies (see Figure 1), at Tc = 150°C
the median life of GaN is 1 × 109 hours vs. 1 × 106 hours for
GaAs. At 1 × 106 hour median life, GaN can operate 75ºC
hotter than GaAs (i.e., 225°C for GaN vs. 150°C for GaAs).2
Overall, GaN can lead to smaller, cheaper, more effi-
cient and higher power RF modules. For example, a single
GaN MMIC high power amplifier (HPA) can replace a pair
of GaAs HPAs which need a power combiner/divider, ex-
tra supporting components and additional assembly. Effi-
ciency is gained not only at the MMIC level, where on-chip
combining losses are reduced, but also at the module level,
where combining losses are eliminated when two or more
GaAs HPAs need to be combined.
Use of GaN HPAs in solid-state phased array radar pro-
vide numerous benefits. As described, GaN supports higher
output power, higher transmit duty factor and longer pulse
lengths than previous technologies, such as GaAs and Si
BJT. These allow smaller aperture sizes and reduce overall
system acquisition costs. Since GaN operates at higher ef-
250 ficiency, operational costs are also reduced, as less prime
power is consumed and less heat is dissipated, reducing
the need for active cooling. An example of this would be
an existing radar which utilizes a solid-state semiconduc-
tor such as GaAs. An equally sized radar employing GaN
would benefit from a significant increase in radar range with
a slight increase in prime power. Conversely, for the same
prime power, the power density of GaN could significant-
ly reduce the radar aperture. This has first order effects of
reducing acquisition costs, including both front-end and
back-end electronics. Lastly, GaN has higher reliability than
previous technologies. Higher reliability reduces operation-
al costs through reduced maintenance and spare parts.
SPACE FENCE RADAR
Space Fence is a ground-based system of S-Band radars
designed to greatly enhance the U.S. Air Force Space Sur-
veillance Network. Space Fence provides unprecedented
sensitivity, coverage and tracking accuracy and contributes
to key space mission threads, with the ability to detect,
track and catalog small objects in low Earth orbit (LEO), me-
dium Earth orbit (MEO) and geosynchronous orbit (GEO).
Space Fence’s capabilities of detecting, tracking and cat-
aloging hundreds of thousands of satellites and debris in
orbit around the earth will revolutionize space situational
awareness.3
Space Fence includes up to two minimally manned radar
sites (see Figure 2). The first radar site is under construction
on Kwajalein Atoll in the Pacific Ocean near the equator
and is expected to become operational in late 2018. The
second site, currently an unfunded contract option, is locat-
ed in Western Australia. The sensor sites provide assured
coverage for objects in LEO and are integrated through an
operations center located in Huntsville, Ala.
The initial Kwajalein radar will provide a persistent sur-
veillance “fence” comprised of thousands of radar beams
covering LEO altitudes. As the Earth rotates, this fence
sweeps the space around the Earth, providing assured cov-
erage to detect satellites and orbital debris. To form high
quality orbital estimates, objects that cross the fence are
tracked over long arcs with dedicated beams. Space Fence
can also be tasked to search for higher altitude objects in
MEO and GEO. The optional second site will complement
the first site’s LEO coverage and also provide tasking capa-
bility to MEO and GEO.3
As shown in Figure 3a, each radar site features a de-
sign with closely spaced but separate transmit and receive
phased array antennas, prime power and liquid cooling.
The transmit array building houses a 36,000 element trans-
mit phased array antenna beneath an air supported low
loss Kevlar environmental radome. The receive building
supports an 86,000 element array, also under a low loss
5
s Fig. 3 Space Fence radar site (a) cutaway of the transmit array (b) and cross-section of the “radar-on-a-board” transmit LRUs (c).

Element-Level vs. Subarrayed DBF Frequency Multiplexing

Sequentially Transmit Multiple RFs Within Receiver Band.
Element-Level DBF: Simultaneously Receive All.
Unconstrained Tx Rx
Instantaneous FoR
North South Scan

1-D Subarrayed DBF:

Instantaneous FoR
Constrained in One Dimension
Time
2-D Subarrayed DBF:
Instantaneous FoR Instantaneous Receiver Band
Constrained in Both Dimensions

Frequency
(b)
(a) East/West Scan f1 f2 f3 f4 f5

Sequential Operation Across Wide Field-of-Regard Sequential Tx and Simultaneous Rx Across Wide FoR

Compress

GaN Needed for Resulting

Long Transmit Pulse and
(c) High Duty Factor

s Fig. 4 Digitizing at the element, instead of combining multiple elements into subarrayed receivers, enables beams to be formed anywhere
in the array FoR (a). Multiple frequency channels, each for a different radar function, can be received at the same time by beams on the receive
array, allowing many functions to be performed simultaneously (b). GaN is the only technology that supports the “machine gun” transmit se-
quence, with long pulses and high duty factor (c).

Kevlar radome. Both arrays are provided power and cool-
ing through the common services building. Radar data pro-
cessing and control of the apertures is performed off-array
in commercial off-the-shelf (COTS) processing equipment
located within the operations building. Both transmit and
receive arrays are automatically calibrated with horns that
are mounted on calibration towers and can transmit or re-
ceive test signals. The extremely large phased arrays are
optimized for high availability and low lifetime support costs
and use GaN HPAs for transmit amplification, providing un-
precedented sensitivity to detect small objects. On receive,
digital beam forming (DBF) at the element level permits
thousands of simultaneous beams instantaneously in any
6
direction. This enables the system to provide persistent
LEO surveillance coverage while simultaneously tracking
hundreds of objects, performing cued search tasks in other
surveillance regimes (including MEO and GEO) and sup-
porting user-defined flexible surveillance volumes. Transmit
and receive arrays are oriented to face straight up and are
designed integrally with the building (see the transmit array
cutaway in Figure 3b). A scalable facility structure supports
liquid cooled cold plates, which house the radar electron-
ics. Radiator tiles are mounted on the top of the cold plates
while “radar-on-a-board” digital transmit and receive line
replaceable units (LRU) are mounted on the sides (Figure
3c). Each transmit LRU incorporates digital waveform gen-
 200

Normalized Performance (%)

180 176%
160 d
iel
140 dY EW-Band Yield
an
120 S-B 124%
100 Cycle Tim
e
80 77%
60
Wafer Process Co 51%
40 st
20
0
Baseline Mid-Point Final
Results Results Results
Program Timeline
(a)

(a) 1E10
1E09
1E08
1E07
1E06

Hours
1E05
1E04
1E03
1E02
1E01
150 200 250 300 350 400
T (°C)
(b)

(b) (c)
s Fig. 6 Wolfspeed DPA Title III performance improvement (a) and
s Fig. 5 Space Fence prototype: antenna building (a) mission op- reliability (b).
erations center (b) and critical design review (CDR) demonstration.

eration, up-conversion to S-Band and high power GaN am-
plification for eight transmit radiating elements. Mounting
the LRUs on the sides of the cold plates provides the GaN
HPAs with a direct and efficient thermal path. To provide
high system availability, the LRUs are serviceable from be-
neath the array and can be removed and replaced in less
than 1½ minutes while the array is operating.
GaN high power amplification was one of the criti-
cal enabling technologies for the Space Fence solution.
Relative to other technologies, the high output power of
GaN reduces the number of transmit elements to achieve
the required sensitivity for the target size, which reduces
overall acquisition cost. GaN’s high efficiency also reduc-
es power consumption and heat dissipated, which reduces
operational costs for the sensor site. In order to effectively
support the LEO orbital regime (and tasking up to GEO)
and get sufficient energy back for detection, transmit pulse
lengths need to be long. Previous technologies, such as
GaAs or Si BJT didn’t support these pulse lengths at the
required output power.
The long pulse capability of GaN in the transmit array
also enables extremely efficient timeline utilization of the
radar when combined with element level DBF in the receive
array. Space Fence has a receiver connected to each array
element within the receive array to digitize the returned
signals. Unlike subarrayed antennas, which combine mul-
tiple elements in microwave electronics prior to digitizing
to reduce the number of receivers, the beams in an ele-
ment level DBF system can be simultaneously placed any-
where in the field-of-regard (FoR) of the array. Subarrayed
approaches limit the digitally formed beams to constrained
volumes and require changing analog phase shifters to
move the volume from one radar event to the next (Figure
4a). Space Fence is able to use its flexibility along with fre-
quency multiplexed functions within the receiver band to
form thousands of beams simultaneously (Figure 4b). This
allows many functions that would have been performed
sequentially to be performed simultaneously, reducing the
Space Fence array sizes along with the associated acquisi-
tion cost and operating costs. Use of GaN HPAs are needed
to support the resulting concatenated “machine gun” like
transmit sequence, which is longer and transmit higher duty
factor than supported by other technologies (Figure 4c).
At the start of concept development for Space Fence in
2007, Lockheed Martin leveraged an existing Independent
Research and Development (IRAD) project that was matur-
ing GaN for use in radar applications. Over the course of
the Space Fence development, GaN was optimized to the
Space Fence application for additional efficiency and re-
duced operational costs. Lockheed Martin also embraced
an open foundry concept and worked with two suppliers to
develop GaN for Space Fence. Space Fence transmit LRUs
have been successfully tested with GaN modules from both
of these companies. Use of multiple suppliers reduced pro-
gram risk and provided competition to reduce acquisition
costs.
With the large Space Fence arrays and number of GaN
devices, high reliability is necessary to keep maintenance
costs down. To ensure mature technology, Lockheed Martin
tested the Space Fence GaN amplifiers at the module, at

7
 700 Other
PtP
VSAT
600 CATV
Infrastructure
Military
500

Revenue ($M)
400

300

200

100

0
(a) 2014 2015 2016 2017 2018 2019
(b)

s Fig. 7 Commercial and military products using GaN (a) and projected GaN market growth (b). Commercial pull for GaN will be more than 2x
military applications by 2019.6

the transmit LRU and at the array. After years of extensive main focus of the GaN Title III program was to improve the
testing (more than 5,000 hours of life testing, including ac- manufacturability of GaN, essentially to increase the MRL
celerated life testing), the technology has proven to be ex- to level 8. The program was structured as three phases: 1)
tremely robust, showing a high reliability confidence level baseline manufacturing readiness, 2) improvement and re-
which supports the rigorous operational availability of the
mission. TABLE 1
Since 2011, Lockheed Martin has had an operational end-
GAN MARKET SHARE STRATEGY ANALYTICS, JANUARY 2015
to-end system prototype employing all its critical technolo-
gies, including GaN and element level DBF, in scaled-down Military Segment Total Worldwide
arrays (see Figure 5). Prototype system data was used by
Wolfspeed 19% 25%
the U.S. Air Force in a technology readiness assessment in
2015, resulting in a Technology Readiness Level (TRL) 7 and Qorvo 24% 22%
Manufacturing Readiness Level (MRL) 7. In January 2016, Sumitomo 17% 22%
an Integration Test Bed (ITB) using final production hard- Raytheon 13% 7%
ware was commissioned in Moorestown, N. J. The ITB sup-
ports hardware/software integration, maintenance training Northrop Grumman 11% 6%
and verification testing; it will provide remote support to UMS 6% 6%
the integration of Sensor Site 1 on Kwajalein Atoll. Others 10% 12%

OPEN FOUNDRY MODEL

The development of GaN with ever increasing govern-
ment funding has been occurring since the 1970s.4 Since
the early 2000s, both the Department of Defense and com-
mercial foundries have considerably increased this fund-
ing. One example is the DARPA Wide Bandgap program
which funded various semiconductor foundries to mature
the technology, investing approximately $150 million. This
program followed the successful initiatives of the 1980s and
1990s to fund GaAs, such as DARPA’s MIMIC and MAFET
programs. Among the key parameters and facets of the
technology development were the physics of the devices,
to understand and unleash the potential of the superior
physical properties of the semiconductors, and the devel-
opment of accurate models.
The completion of the DARPA Wide Bandgap program
in 2011 led to an additional program funded through the
Defense Production Act (DPA) Title III program office. The
mission of the DPA Title III program is to “create assured,
affordable and commercially viable production capabilities
and capacities for items essential for national defense.” The
8
finement of processes and 3) final manufacturing readiness
assessment.
Wolfspeed was able to substantially increase their yield
over the three phases, while simultaneously reducing cycle
time and wafer processing costs (see Figure 6).5 Over the
three phases, yield increased 76 percent for S-Band de-
vices, cycle time reduced 23 percent and wafer costs de-
creased 49 percent. Reliability of 1 × 107 hours was demon-
strated at a channel temperature of approximately 275°C.
Through manufacturing process refinement and building
significant robust wafer lots, Wolfspeed reached MRL 8 in
2014, i.e., demonstrated pilot line capability and ready to
begin low rate initial production (LRIP). One of the other
major commercial foundries, Qorvo (formerly TriQuint) was
the only funded foundry that achieved MRL 9 in 2014, i.e.,
demonstrated LRIP and capability in place to begin full rate
production (FRP).
GaN has been inserted into several Lockheed Martin
and other DoD contractor systems and has achieved TRLs
through level 9. Space Fence has achieved TRL 7. Since the
completion of both the DARPA Wide Bandgap and DPA Ti-
tle III programs, GaN has been following a Si CMOS trajec-
tory, in terms of applications and utilization in commercial
markets.6 GaN has found its way into base stations, med-
ical equipment and even residential/commercial lighting
available at local hardware stores (see Figure 7).
Lockheed Martin employs an open foundry model (LM
OpenGaN), which leverages the commercial market for the
best technology at the most competitive cost. The cost of
owning and running a foundry year after year can be sig-
nificant. Given the growth projections for semiconductors,
in general, it makes more sense to allow high commercial
volumes to mature the technology process and drive costs
down.7 A study conducted by Strategy Analytics6 (see
Table 1), identified Qorvo and Wolfspeed as GaN leaders,
with no other domestic suppliers close to providing equiv-
alent volume. In the semiconductor industry, volume pro-
duction lends itself to higher maturity, stable processes and
lower costs.
Lockheed Martin adopted an open foundry model to
address the Department of Defense’s official memorandum
on Better Buying Power 3.0 and the “3rd Offset,” with the
goal of increasing the capability and rapid technology de-
velopment required by the warfighter.8 It is not economical
to have any particular foundry develop various semicon-
ductor processes, not just III/IV (i.e., GaAs, GaN, InP) but
also SiGe and CMOS.
GaN has had significant investment over the past 10
years and is proven to be field ready, reliable and cost com-
petitive with existing technologies (e.g., GaAs). The system
benefits enable enhanced capabilities for backfit military
systems and are required for the performance of future sys-
tems. The Space Fence program would not have been pos-
sible without GaN nor as successful without the affordability
enabled by the LM Open GaN foundry model.
THE FUTURE
The next generation of GaN development is focused
around increasing capability by pushing into higher oper-
ating frequencies, improving thermal performance and en-
abling chip-scale integration of GaN with other IC technol-
ogies. Developments of 150 nm and 90 nm process nodes
will extend ft to 60 and 100 GHz, respectively.
While DBF systems such as Space Fence represent a
significant capability upgrade, the challenges from the in-
creased IC design complexity in the areas of size, weight
and power (SWAP), bandwidth and latency performance
must be overcome. To address this, programs such as DAR-
PA’s Diverse Accessible Heterogeneous Integration (DAHI)
program are focused on developing chip-scale integration
of GaN with high density Si CMOS, as well as other technol-
ogies such as InP and MEMS. These resulting capabilities
will enable the wider proliferation of the high performance
mixed-signal integration solutions required to develop the
capabilities to further advance state-of-the-art sensor sys-
tems.
While GaN is capable of generating extremely high RF
power densities, thermal management remains a significant
challenge, especially at higher frequencies where thermal
density is most extreme. On the DARPA IceCool program,
Lockheed Martin has made tremendous gains in unlocking
the ultimate potential of millimeter wave GaN by devel-
9
oping a micro-fluidically cooled HPA with a 3×reduction in
thermal resistance compared to conventional thermal man-
agement solutions. This IceCool solution enables an 8.3 dB
increase in output power for the same device, while simul-
taneously reducing operating temperature by increasing
power-added efficiency between 2.5× and 3.5×.9 Overall,
these advancements will further extend the true potential
of GaN and its ability to realize tremendous capability up-
grades for a wide variety of systems. ■
ACKNOWLEDGMENT
The authors would like to acknowledge and thank the
Space Fence and ICECool sponsors (AFLCMC, DARPA and
AFRL) and partners for their financial and technical support.
References
1. Liou Schwierz, “Modern Microwave Transistors,” Wiley, 2003.
2. R. Hall, “GaN at Qorvo” Presentation, August 2015.
3. Joseph A. Haimerl and Gregory P. Fonder, “Space Fence System
Overview,” Advanced Maui Optical and Space Surveillance Technolo-
gies Conference Proceedings, September 2015.
4. M. Yoder, “Gallium Nitride Past, Present, and Future,” IEEE/Cornell
Conference on Advanced Concepts, 1997.
5. Ryan Fury, Scott T. Sheppard, Jeffrey B. Barner, Bill Pribble, Jeremy
Fisher, Donald A. Gajewski, Fabian Radulescu, Helmut Hagleitner,
Dan Namishia, Zoltan Ring, Jennifer Gao, Sangmin Lee, Brian Fetzer,
Rick McFarland, Jim Milligan and John Palmour, “GaN on SiC MMIC
Production for S-Band and EW-Band Applications,” CS ManTech Con-
ference, May 13–16, 2013, New Orleans, La.
6. Eric Higham, Strategy Analytics, “An Overview of GaN RF Device
Market,” GaN Roundtable Session, August 26, 2015.
7. Justin Gallagher, “The Commercialization of GaN,” Defense Manufac-
turing Conference Proceedings, November 2015.
8. Under Secretary of Defense, “Implementation Directive for Better
Buying Power 3.0 – Achieving Dominant Capabilities through Techni-
cal Excellence and Innovation,” Memorandum, 2015 http://bbp.dau.
mil/docs/BBP3.0ImplementationGuidanceMemorandumforRelease.
pdf.
9. J. Ditri, R. Cadotte, D. Fetterolf and M. McNulty, “Impact of Microflu-
idic Cooling on High Power Amplifier RF Performance,” Proceedings
of the 15th Intersociety Conference on Thermal and Thermomechani-
cal Phenomena in Electronic Systems (ITHERM 2016), Las Vegas, Nev.,
2016.
Solid-State PAs Battle TWTAs for
ECM Systems
Rick Montgomery and Patrick Courtney
Qorvo, Greensboro, N.C.
E
lectronic countermeasures or ECM sys-
tems are typically comprised of receivers,
processors, displays and jamming trans-
mitters. Until recently, solid-state amplifiers fell
short of the required combination of power,
bandwidth and efficiency for the transmitter of
an ECM system. Maturating GaN power ampli-
fier MMICs and low loss, broadband combining
techniques now make it possible to meet the
power, bandwidth and efficiency requirements
of ECM systems with solid-state power ampli-
fiers (SSPA). Compared to GaAs and other sol-
id-state semiconductor materials, GaN provides
an order-of-magnitude increase in transistor
power density, and the higher impedance of the
devices eases the design of matching networks.
Historically, traveling wave tubes (TWT) and
other vacuum tubes have provided the micro-
wave power for ECM transmitters. Since the
1950s, the broadband, high-power microwave
amplification necessary for ECM transmitters
was only feasible with vacuum tube technolo-
gy and, in particular, with traveling wave tube
amplifiers (TWTA). ECM jamming transmitters
typically need to generate hundreds of watts of
microwave power over several octaves. The effi-
ciency of the amplifiers must be high enough to
meet the limited power budget of airborne plat-
WWW.MWJOURNAL.COM/ARTICLES/28478
10
forms and the heat generated can be dissipat-
ed. TWTAs were the only technology that could
meet these critical requirements.
SOLID-STATE VS. TUBE
Solid-state devices have long been preferred
to vacuum devices. Vacuum tubes, with their as-
sociated high voltage power supplies—typically
in the multi-kV range—have lower reliability than
solid-state devices operating with low voltage
power supplies (i.e., under 50 V). Manufactur-
ers and users of vacuum tubes face diminishing
sources of supply and material shortages.
Solid-state devices generate lower noise and
have better linearity than vacuum tubes. For in-
stance, solid-state devices in “standby mode”—
where the DC bias voltage is applied with no RF
input signal—generate significantly less noise
power across the spectrum. Noise figures for a
medium power TWT can be around 30 dB, ver-
sus about 10 dB for a solid-state GaN MMIC PA.
This is a significant difference in an ECM system,
as the lower noise may allow the transmitter’s
output stage to remain in standby mode when
not transmitting. The overall switching time de-
creases, since the main DC power to the PA does
not need to be switched on and off.
 43
–40°C +25°C +85°C
42
41
Pout (dBm)
40
39
38
37
36
35
0 1 2 3 4 5 6 7 8 9 Frequency (GHz)
Frequency (GHz)
s Fig. 1 Output power vs. frequency and
temperature of a Qorvo QPA1003P GaN
MMIC, with 15 dBm CW input power, 28 V
bias and 650 mA current consumption.
Another operational benefit of a solid-state transmitter
is the reduced harmonic content in the output signal. A sol-
id-state PA that operates over an octave or greater band-
width will typically have worst-case harmonic content about
8 dB down from the fundamental at its saturated output
power. The harmonic content of a vacuum tube will only be
down 2 dB from the fundamental under the same operat-
ing conditions. These higher harmonics can impose stricter
filtering requirements for the transmitter, driving larger and
costlier components for the overall ECM system.
POWERING UP WITH GaN
While GaN devices have significantly increased the pow-
er density, power and bandwidth over other heterojunction
semiconductor technologies, a single device or MMIC still
has insufficient power for most ECM system transmitters. It
is not unusual to have a requirement of 100 W or more from
2 to 7.5 GHz. Figure 1 shows the output of a single Qorvo
GaN power MMIC. This packaged MMIC nominally delivers
10 W from 1 to 8 GHz, but the output power decreases at
85°C backside temperature to as low as 8 W. More than 10
of these MMICs must be combined to deliver 100 W across
the band and over the temperature range required in an
ECM system.
There are many ways to power combine devices to
make a SSPA. For an ECM system transmitter, the ap-
proach must have low loss and wide bandwidth. Many
combining techniques use two-port binary combiners
such as Wilkinson or magic tees. Combining two MMICs
requires a single two-port combiner, four MMICs re-
quires three combiners and 16-way combining requires
15 combining elements. Magic tees have relatively low
loss; however, they typically operate over a maximum of
10 percent bandwidth, and double-ridged magic tees
s Fig. 4Spatium amplifier with clamp to conduct heat from the
MMIC PAs.
Input
Antipodal 54
Finline MMIC Output Antipodal 53 –45°C +25°C +71°C
Antennas Assemblies Finline Antennas
52
Pout (dBm)
51
50
49
48
47
2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5
Input Output
Transition Transition
s Fig. 2 Spatium amplifier structure. s Fig. 3 Measured output power of the
Spatium amplifier that combines 16
QPA1003P MMICs.
only have about an octave bandwidth, which is short of
a 2 to 7.5 GHz ECM requirement. With two-way com-
bining, four stages of combining are needed to achieve
the desired power. A typical double-ridge magic tee
has 0.3 dB of loss at these frequencies, so the total loss
through the combiner would be 1.2 dB. Combining the
30 percent efficient GaN PA MMICs shown in Figure 1
through a 16-way magic tee, the efficiency of the com-
bined output would be about 23 percent and deliver
approximately 95 W output at 6 GHz at 85°C. However,
the typical double-ridge magic tee network only works
over an octave of bandwidth (e.g., from 2 to 4 GHz).
SPATIAL COMBINING
Spatial combining techniques are potentially lower loss
than circuit-based techniques. Spatium® is Qorvo’s patent-
ed coaxial spatial method of power combining (see Fig-
ure 2). It uses broadband antipodal finline antennas as the
launch to and from the coaxial mode, splitting into multiple
microstrip circuits, then combining the power from those
circuits after amplification with a power MMIC. It uniquely
enables broadband, efficient and compact combining of
multiple power MMICs in a single combining step, with free
space as the combining medium. A typical Spatium design
combines 16 devices in one step, with a combining loss of
only 0.5 dB.
Combining 16 of the MMICs from Figure 1 yields an SSPA
efficiency of 27 percent, compared to 30 percent for each
MMIC. This is a significant difference compared to the 23
percent achieved with magic tee combining. The increased
combining efficiency enables higher output power from a
given prime power as well as reducing the heat dissipation.
An actual Spatium amplifier was designed that com-
bined 16 radial blades with the Qorvo GaN MMIC PA on
each blade. Figure 3 shows the measured output power
versus clamp surface temperature; the baseplate tempera-
ture below the MMIC is approximately 12°C hotter than the
clamp temperature, so the maximum baseplate tempera-
ture is 85°C. This unit achieves greater than 100 W from 2
to 7.5 GHz and an average efficiency of 25 percent.
THERMAL DESIGN
Thermal management is one of the design challenges
when using a solid-state amplifier in an ECM transmitter. In
a typical application, the outer surface of the clamp around
the Spatium SSPA is conduction cooled on one or more
sides (see Figure 4). For some systems, a liquid coolant
may be available, for others a heat sink with fans. The clamp
11
 106
MTBF = (1)
16.5 × (1.00001)P × (1.1)F

where P is the rated power in watts, from 1 mW to 40 kW,

and F is the operating frequency in GHz, from 100 MHz to
18 GHz. Utilizing this formula, a TWT with an output power
of 150 W at a frequency of 7.5 GHz has an MTBF of 29,609
hours. This is 2.5x lower than that of the comparable sol-
id-state Spatium power amplifier module under similar en-
vironmental conditions.

SUMMARY
For the first time, GaN MMICs and broadband spatial
s Fig. 5 Thermal simulation of the Spatium SSPA, showing cross combining techniques such as Spatium allow ECM system
section of the construction. designers to use reliable, solid-state amplifiers instead of
is designed to make contact with all of the blades in the TWTAs. The ability to deliver hundreds of watts over broad
Spatium and provide a conduction path to a cold plate or bandwidths, while staying within the prime power available
heat sink. Spatium blades and clamps can be made of dif- from the platform and dissipating the heat to ensure reli-
ferent metals, including aluminum and copper. Size, weight able operation, opens up new system opportunities for a
and power trade-offs determine the appropriate material solid-state ECM transmitter. Table 1 shows the frequency,
set for a given application. power and efficiency achieved with three recent Spatium
The thermal resistance from the back of the MMICs to amplifiers. The size and weight of these SSPAs are less than
the mounting plates can be calculated and used to derive the boxes previously occupied by their respective TWTAs.■
the backside MMIC temperature. From the thermal resis-
tance of the MMIC and package, the junction temperature
of the MMIC can be calculated and, from that, the reliability
of the SSPA can be estimated. Figure 5 is a thermal simula-
tion of the SSPA shown in Figure 4, with the MMICs operat-
ing at saturated output power and the worst-case efficiency
within the frequency band (i.e., 25 W dissipated per MMIC).
The thermal model shows an approximate 12°C rise from
the coolest spot of the outside surface of the clamp to the
backside of the packaged MMIC and an additional 164°C
temperature rise from the back of the package to the junc-
tion of the output transistor, assuming 6.56°C/W thermal
resistance. The junction temperature of the MMIC is esti-
mated to be 247°C with the surface of the clamp held to
71°C. At 247°C junction temperature, the MTBF of the
MMIC is some 1.2 million hours.
The MTBF of the overall Spatium module will be the
MTBF of the individual MMICs divided by the number
of MMICs: 75,000 hours. The calculation deems a single
MMIC failure to be a failure of the entire amplifier assem-
bly—a worst-case assumption since the Spatium amplifier
will gracefully degrade with the failure of any individual
MMIC (i.e., approximately 0.7 dB reduction in output pow-
er per MMIC failure).
For a TWT, MIL-HDBK-217F Notice 2 provides the fol-
lowing formula to calculate the MTBF in a ground fixed en-
vironment:

TABLE 1
DEMONSTRATED SPATIUM SSPA PERFORMANCE
Frequency Range Nominal Output Nominal Power-
(GHz) Power (W) Added Efficiency
2 to 6 300 30%
2 to 8 150 25%
2 to 18 60 15%

12
 Yes we GaN
Boonton RF/microwave benchtop power meters and USB power sensors are ideal for
testing GaN amplifier performance with best-in-class rise time, acquisition speed,
time resolution, and usability to enable a wide variety of measurements, including:

• Gain • Linearity • Reflected power

• P1 dB compression • Dynamic range • Stability over time

Utilize the industry-best rise time (<3 ns) and time

resolution (100 ps) to characterize even the most
demanding radar and communication signals.

Unique Real-Time Power Processing™ enables up to

100,000 measurements per second with virtually
no missed pulses or signal anomalies.

With 16 automated pulse measurements as well

as various marker measurements, including droop,
characterizing the performance of GaN amplifiers
has never been easier.

Leaders in Power Measurement for more than 70 Years

www.boonton.com
GaN Powers Microwave
Point-to-Point Radios
Kristoffer Andersson, David Gustafsson and Jonas Hansryd
Ericsson, Gothenburg, Sweden

Gallium arsenide (GaAs) has been the key technology enabler for high performance power amplifiers
and receiver low noise amplifiers in the microwave frequency bands; however, the maturity of gallium
nitride (GaN) technology provides an opportunity for substantial system level improvements. The high
dielectric breakdown field of GaN results in higher voltage operation and hence higher power densities.
With the help of GaN, the efficiency and the maximum output power of a microwave radio can be
substantially increased — directly impacting the network cost of ownership. Today, GaN monolithic
microwave integrated circuits (MMIC) are being introduced in commercial microwave radios.

T
oday there are as many mobile sub-
scriptions as there are people in the
world and the number of subscribers,
which now includes machines and
things, is rising at an increasing rate. High perfor-
mance microwave radios have been key enablers
in achieving the successful rollout of mobile net-
works. By 2020 it is expected that 65 percent of
the radio base station sites (excluding China, Ja-
pan, Taiwan and South Korea) will be connected
to a core network via a microwave radio.1
The major advantages of microwave radios
are fast time-to-market and low cost of owner-
ship. Two important parameters influencing cost
of ownership are energy efficiency and system
gain. Increasing the system gain allows for the
use of smaller antennas while maintaining ca-
pacity and link availability. Cost of ownership is
typically dominated by site rental which often
represents 30 to 50 percent of the total cost.2 A
small antenna size results in a lower rental cost
due to reduced wind loading of the mast. This
also leaves room for additional radio links.
There are a number of components in the
radio system enabling improved system perfor-
mance such as high gain/high efficiency pow-
er amplifiers (PA), low noise receiver amplifiers
(LNA) and error correction codes that regenerate
incoming signals and mitigate disturbances such
as interference, noise and spurious.
The available frequency bands for microwave
radio span from 6 to 70/80 GHz with a total avail-
able bandwidth of about 40 GHz. For the tradi-
tional microwave bands (6 to 42 GHz), available
channel bandwidths vary from a few MHz up
to 112 MHz at the highest carrier frequencies.
At 70/80 GHz, the channel bandwidth is wider
than 1 GHz. To use the available spectrum in the
most efficient way, modern radio systems often
support modulation formats up to 4096-QAM.
Figure 1 shows the use of different frequency
bands for microwave radio applications around
the world. The most used bands are between 15
and 23 GHz.
GaN MMIC TECHNOLOGY
GaN comes with the distinct advantages of
high output power density and high voltage
operation. From a power amplifier perspective,
GaN is a game changer. Using the same size
transistors, it is possible to boost output power
by at least four times compared to GaAs. Higher

WWW.MWJOURNAL.COM/ARTICLES/27778
14
 Frequency Band (GHz)
Region CMOS Bulk, CMOS SOI, SiGe HBT
6 7 8 10 11 13 15 18 23 26 28 32 38 42 60 70/80 GaN HEMT
North America GaAs mHEMT, GaAs pHEMT
Latin America
103
Sub Saharan
Africa 102

Output Power (W)

Mediterranean 101
Western &
Central Europe 100
Northern Europe 10–1
& Central Asia
Middle East 10–2
India 10–3
South East Asia 10–4 0
& Oceania 10 101 102 103
North East Asia Frequency (GHz)

s Fig. 1 Microwave radio market share vs. frequency band vs. geographic region.3
s Fig. 2 Published power amplifier output
power, reflecting capability of silicon, GaAs and
GaN technologies.

power density means less complex circuitry, hence higher
efficiency and wider bandwidth. Several semiconductor
companies now have access to GaN MMIC technology
offering high performance processes with gate lengths
less than 150 nm. This enables high performance MMICs
to greater than 40 GHz. Figure 2 shows the potential of
GaN power amplifiers compared to other semiconductor
technologies. The figure shows regions of maximum output
power versus frequency for published power amplifiers us-
ing three different semiconductor technologies (GaN, GaAs
and silicon CMOS/SiGe). The dashed line shows a drop in
power capability as a function of frequency for all technol-
ogies.
Existing barriers to wide-scale adoption of GaN technol-
ogy in microwave point-to-point links is wafer pricing, pack-
aging and memory effects due to trapping. The current
generation of GaN MMICs is manufactured on semi-insu-
lating silicon carbide (SiC) substrates. This technology pro-
vides the highest performance but not the most attractive
price. Today the industry is moving to 6-inch SiC substrates,
which helps push cost towards similar levels as GaAs. To
further reduce cost, several companies are focusing on
GaN-on-Silicon (GaN-on-Si) technology. GaN-on-Si is not
as mature as GaN-on-SiC, however, and the performance
will not be as high. One reason is the lower thermal con-
ductivity of Si compared to SiC.
For telecommunication equipment, organic overmold
packaging is the dominant technology due to its low cost
and high reliability. However, only a very limited number
of suppliers currently offer GaN MMICs with the necessary
surface passivation to support organic overmold packag-
ing.
Trapping in the GaN transistor may result in dynamic ef-
fects with slow time constants in the range of milliseconds
to seconds.19 The effects can
appear as abrupt changes in
small-signal gain following
TYPICAL PERFORMANCE OF PACKAGED GaAs CLASS AB
high output power events.
This is more of an issue for
radio base stations having Reference Frequency (GHz)
Saturated
Output
very bursty traffic patterns Power (dBm)
than for microwave radio
4 7
links where output power
changes less rapidly. 5 23
6 38
MICROWAVE RADIO POWER AMPLIFIERS
A power amplifier is at maximum efficiency when operat-
ing at peak output power, while lowering the output power
will significantly reduce efficiency. For linearity, the converse
is true; the most linear operation is achieved at low output
power. To mitigate the impairments associated with high
efficiency operation, digital predistortion (DPD) techniques
are used to compensate for power amplifier nonlinearities.
Power amplifiers for radio base stations are typically
optimized for efficiency; power-added efficiencies as high
as 50 to 70 percent are not uncommon. Linearity is then
recovered using DPD techniques. This focus on efficiency
is motivated by the high output powers (100 W average)
of these base station systems. Even a moderate increase
in power-added efficiency leads to a significant reduction
in power consumption and cooling requirements.
Power amplifiers in microwave radios for point-to-point
communications are operated Class AB to support high or-
der modulation schemes (up to 4096-QAM). If a DPD sys-
tem is available, it is typically used to enable high order
modulation rather than for boosting efficiency. The maxi-
mum saturated output power for a GaAs MMIC amplifier is
about 40 dBm up to 10 GHz, falling to about 26 dBm at 86
GHz (E-Band). The associated maximum efficiency is about
35 percent up to 10 GHz and about 15 percent at 86 GHz.
For linear operation (i.e. 10 dB from saturation), these effi-
ciencies drop rapidly to below 10% (3% to 10% depending
on frequency). These power levels represent what is possi-
ble to achieve in a commercial setting. Striving for higher
output power increases cost and power consumption.
For most point-to-point frequency bands, there are a
number of organic packaged GaAs MMIC power amplifi-
ers. For the lower frequencies, saturated output power typ-
ically lies in the range of 2 to 4 W and the power-added
TABLE 1
POWER AMPLIFIERS
P1dB
(dBm)
Third Order
Intercept Point
Power-Added Supply
Efficiency for Linear Voltage (V)
(dBm) Operation (%)
35 34 43 <5 6
33 31 38 < 3 6
28 27 38 <2 4

15
 TABLE 2 power increases. This load
modulation ensures that the
PERFORMANCE OF GaN DOHERTY POWER AMPLIFIERS ABOVE 6 GHz carrier amplifier operates
Reference Frequency Saturated Power-Added Supply Physical near saturation (and hence
(GHz) Output Power Efficiency for Linear Voltage (V) Configuration at high efficiency) over a
(dBm) Operation (%) wide range of output pow-
ers — typically 6 dB or more.
9 7 36 31 - 39 28 Bare Die
Although the theoretical
11 7 35 35 20 Bare Die Doherty amplifier is a linear
14 7 38 36 28 Bare Die amplifier, the practical reali-
17 7 42 21 28 Packaged zation is inevitably very non-
linear. Hence, the radio base
15 10 36 34 20 Bare Die
station is also equipped with
16 23 37 27 20 Bare Die a digital predistortion sys-
tem that provides linear op-
Vg1 Vg2 Vd eration.
Given the success of the Doherty amplifier in the radio
base station market, there have been attempts to carry over
the Doherty technology to the point-to-point segment. The
principal circuit level challenges have been the shorter
wavelengths in the point-to-point communication bands
and the low power density of GaAs. Shorter wavelengths
1.5 mm

necessitate an integrated solution in contrast to the sub-6

GHz Doherty amplifiers, where the combiner and splitting
networks are implemented on the amplifier printed circuit
RFin RFout board. The low power density of GaAs puts a limit on the
maximum output that can be achieved. With the advent of
short gate length GaN MMIC technology, it is finally pos-
2.1 mm sible to realize Doherty power amplifiers beyond 6 GHz,
delivering more than 10 W in saturation.
s Fig. 3 7 GHz GaN MMIC Doherty amplifier.11 Recent work with GaN Doherty power amplifiers operat-
ing primarily in the 7 GHz bands9-15 and at 23 GHz16 show
Power-Added Efficiency (%)

60 that state-of-the-art GaN technology is capable of achiev-

50
6.8 GHz 7.2 GHz
7.6 GHz 8.0 GHz
40
30
20
10
18 20 22 24 26 28 30 32 34 36 38
Output Power (dBm)
s Fig. 4 Power-added efficiency of 7 GHz GaN MMIC Doherty
amplifier.11
efficiency for linear operation is slightly less than 5 percent.
Table 1 shows the typical performance of commercially
available packaged GaAs Class AB power amplifiers. Taken
from vendor datasheets, it serves as an illustration of the
current state for point-to-point amplifiers: linear operation
at the expense of efficiency.
Efficient Point-to-Point Communications Power Amplifiers
Wireless radio base stations have for some time been
equipped with Doherty power amplifiers that provide high
efficiency in both saturation and in back-off.7,8 In its most
basic form, the Doherty amplifier improves efficiency in
back-off by using two parallel amplifier branches: a carrier
amplifier and a peaking amplifier. The carrier amplifier is
connected to the peaking amplifier via an impedance in-
verter and the peaking amplifier is then connected to the
load. This arrangement at low power presents a high im-
pedance to the carrier amplifier, which decreases as the
16
ing a power-added efficiency of 30 percent (with 8 to 9 dB
back-off) and more than 20 W of saturated output power
(see Table 2). In 2012, Gustafsson et al.,11 demonstrated
the performance of a GaN MMIC Doherty power amplifier
operating in a microwave radio band (see Figure 3). Figure
4 shows its efficiency. Using a DPD system, the amplifier
achieved an average efficiency of greater than 35 percent
while keeping its adjacent channel leakage power ratio
(ACLR) below ‐48 dB for a 10 MHz 256-QAM signal.
More recently, Gustafsson et al.,17 demonstrated the
performance of a GaN hybrid Doherty amplifier. The hybrid
approach with passive GaAs for matching circuits limits the
GaN content, which has the potential to lower the cost of
the Doherty amplifier. The two-stage hybrid amplifier con-
sists of an integrated passive GaAs input, interstage and
output matching circuits and GaN power bars for the ampli-
s Fig. 5 GaN hybrid power
amplifier before overmold
application.17 The power
transistors are GaN HEMTs, and
the passive input, interstage
and output networks are
fabricated in GaAs.
 fier stages. Figures 5
and 6 show the in-
side of the package
and the packaged
amplifier, respec-
tively. Using a DPD
system, the amplifier
delivered 34 dBm av-
erage output power
with a 50 MHz 4096-
QAM signal at an av-
erage power-added
efficiency of 20 per-
cent, while fulfilling
Class 8B ETSI spec-
tral requirements18
s Fig. 6 Packaged GaN hybrid power (see Figure 7). Lin-
amplifier mounted to an evaluation earization was per-
board.17 formed using direct
learning table DPD
(24 tables, 16 bins)
Normalized Spectrum (dB)
0
Without DPD
with least square es-
With DPD
–20 ETSI Mask timation of the table
coefficients.
–40
This demonstra-
–60 tion shows that GaN
and Doherty opera-
–80 tion can increase the
–400 –200 0 200 400
Frequency (MHz) efficiency well be-
s Fig. 7 Measured output spectrum of yond able
what is achiev-
in GaAs. As pre-
a GaN hybrid power amplifier, with and
without DPD, driven by a 50 MHz, viously mentioned,
4096-QAM input at 6.65 GHz.17 the Doherty ampli-
fier must be aug-
mented with DPD. The wide bandwidth of point-to-point
channels—up to 112 MHz for the traditional microwave
bands —puts tough requirements on the digital predistor-
tion system. To compensate for seventh-order nonlinear-
ities, digital-to-analog converters should run with at least
7x oversampling (i.e., > 784 MHz). The same goes for the
analog-to-digital converter in the local observation receiver
(>784 MSPS). In addition to high data rates, vertical reso-
40
35
30
Required DC Power (W)
25
20 GaAs
15
10 GaN
5 26
0
0 10
Amplifier Efficiency (%)
s Fig. 8 DC power consumption vs. amplifier efficiency and
average output power. Circled regions show typical performance of
GaAs Class AB, GaN Class AB and GaN Doherty power amplifiers.
lution should be high enough to support very high modu-
lation rates (i.e., 4096-QAM). These data converters have
been historically power hungry and, thus, the efficiency of
the whole radio system is compromised. By using state-of-
the-art data converters, however, it is now feasible to de-
sign digital predistortion systems with low enough power
consumption.
The Promise of GaN PAs in Microwave Point-to-Point
Radios
There are two main uses for GaN in commercial micro-
wave radios. The first is to leverage power density to in-
crease output power and, thus, system gain. Using short
gate length GaN MMIC processes (e.g., 0.15 µm), it is
possible to substantially increase output power, compared
to GaAs, up to at least 23 GHz. This comes at the cost of
increased power consumption, as amplifier efficiencies are
expected to improve marginally. The other use is to im-
prove efficiency by introducing high efficiency power am-
plifier architectures (e.g., Doherty). With a high efficiency
amplifier, it is possible to increase output power while also
decreasing power consumption. Furthermore, the thermal
footprint of the amplifier is reduced, enabling simpler cool-
ing solutions.
Figure 8 shows the required DC power as a function of
power amplifier efficiency for three different average pow-
er levels (30, 33 and 36 dBm). The circled areas indicate
typical performance regions for GaAs Class AB, GaN Class
AB and GaN Doherty power amplifiers. By replacing GaAs
with GaN, an incremental increase in efficiency is achieved.
At these relatively low efficiencies, however, even an in-
cremental increase leads to substantial power savings. An
increase from 5 to 10 percent power-added efficiency can
reduce power consumption by 10 W and, hence, lower the
total cost of ownership. The use of GaN Doherty power
amplifiers enables power-added efficiencies in the range
of 25 percent. This not only decreases power consumption
but also increases system gain and output power.
Looking at GaN opportunities in terms of carrier fre-
quency, there are two distinct regions (see Figure 9). The
first is at lower frequencies (up to about 13 GHz), where
efficient Doherty operation has been demonstrated. Here
one can take advantage of increased efficiency and re-
duced energy consumption or increased output power
without increased power consumption or an enlarged ther-
Pavg = 30 dBm mal footprint. This region utilizes the full potential of GaN
Pavg = 33 dBm
Pavg = 36 dBm technology. The gray area indicates a transition region
Average Output Power (dBm)
38
GaN Power Limit
36 GaN Doherty
34 GaN Class AB
32
30
GaN 28
Doherty
24
20 30 40 50 10 15 20 25 30 35 40
Frequency (GHz)
s Fig. 9 Operating regions for Doherty and Class AB GaN power
amplifiers.
17
between Doherty and Class AB power amplifiers; which
technology is best to use depends on the efficiencies that
can be achieved from Doherty amplifiers above 15 GHz.
For the higher frequency bands (18 GHz and above), it is
now possible to replace existing GaAs power amplifiers
with high power GaN versions to increase output power.
CONCLUSION
GaN technology is now being introduced in commercial
microwave radios. While the first wave of products pro-
vides only an incremental reduction in power consumption,
Doherty amplifiers at microwave frequencies enabled by
GaN and high performance data conversion technology
will, in the coming years, provide power and efficiency im-
provements, just as they have for the base station market.
■ F. Giannini, R. Giofre, L. Piazzon, T. Emanuelsson and T. Wegeland,
References
1. Ericsson, “Microwave Towards 2020- 2015 Edition,” September 2015.
2. Ericsson, “Microwave Towards 2020- 2014 Edition,” 2014.
3. J. Edstam, “Microwave Backhaul Gets a Boost with Multiband,” Erics-
son Technology Review, January 2016.
4. “TGA2701-SM Datasheet,” www.triquint.com/products/p/TGA2701-
SM.
5. “CHA5356-QGG Datasheet,” http://module-csums.cognix-systems.
com/telechargement/9-35-1.pdf.
6. “MAAP-011170 Datasheet,” http://cdn.macom.com/datasheets/
MAAP-011170_V3.pdf.
7. R. Pengelly, C. Fager and M. Ozen, “Doherty’s Legacy: A History of
the Doherty Power Amplifier from 1936 to the Present Day,” IEEE Mi-
crowave Magazine, Vol. 17, No. 2, February 2016, pp. 41–58.
8. V. Camarchia, M. Pirola, R. Quaglia, S. Jee, Y. Cho and B. Kim, “The
Doherty Power Amplifier: Review of Recent Solutions and Trends,”
IEEE Transactions on Microwave Theory and Techniques, Vol. 63, No.
2, February 2015, pp. 559–571.
9. D. Gustafsson, J. C. Cahuana, D. Kuylenstierna, I. Angelov and C.
Fager, “A GaN MMIC Modified Doherty PA with Large Bandwidth and
Reconfigurable Efficiency,” IEEE Transactions on Microwave Theory
18
and Techniques, Vol. 62, No. 12, December 2014, pp. 3006–3016.
10. D. Gustafsson, C. M. Andersson and C. Fager, “A Modified Doherty
Power Amplifier with Extended Bandwidth and Reconfigurable Effi-
ciency,” IEEE Transactions on Microwave Theory and Techniques, Vol.
61, No. 1, January 2013, pp. 533–542.
11. D. Gustafsson, J. C. Cahuana, D. Kuylenstierna, I. Angelov, N. Rors-
man and C. Fager, “A Wideband and Compact GaN MMIC Doherty
Amplifier for Microwave Link Applications,” IEEE Transactions on Mi-
crowave Theory and Techniques, Vol. 61, No. 2, February 2013, pp.
922–930.
12. R. Quaglia, V. Camarchia, M. Pirola, J. J. M. Rubio and G. Ghione,
“Linear GaN MMIC Combined Power Amplifiers for 7-GHz Microwave
Backhaul,” IEEE Transactions on Microwave Theory and Techniques,
Vol. 62, No. 11, November 2014, pp. 2700–2710.
13. V. Camarchia, J. Fang, J. M. Rubio, M. Pirola and R. Quaglia, “7 GHz
MMIC GaN Doherty Power Amplifier with 47 Percent Efficiency at 7
dB Output Back-Off,” IEEE Microwave and Wireless Components Let-
ters, Vol. 23, No. 1, January 2013, pp. 34–36.
14. V. Camarchia, J. J. M. Rubio, M. Pirola, R. Quaglia, P. Colantonio,
“High-Efficiency 7 GHz Doherty GaN MMIC Power Amplifiers for Mi-
crowave Backhaul Radio Links,” IEEE Transactions on Electron Devic-
es, Vol. 60, No. 10, October 2013, pp. 3592–3595.
15. M. Coffey, P. MomenRoodaki, A. Zai and Z. Popovic, “A 4.2-W 10-GHz
GaN MMIC Doherty Power Amplifier,” IEEE Compound Semiconduc-
tor Integrated Circuit Symposium, October 2015, pp. 1–4.
16. C. F. Campbell, K. Tran, M. Y. Kao and S. Nayak, “A K-Band 5W
Doherty Amplifier MMIC Utilizing 0.15μm GaN on SiC HEMT Technol-
ogy,” IEEE Compound Semiconductor Integrated Circuit Symposium,
October 2012, pp. 1–4.
17. D. Gustafsson, K. Andersson, A. Leidenhed, A. Rhodin and T. Wege-
land, “A Packaged Hybrid Doherty PA for Microwave Links,” Europe-
an Microwave Week, October 2016.
18. “ETSI-EN 302 217-2-2 V2.2.1 (2014-04),” www.etsi.org/de-
liver/etsi_en/302200_302299/3022170202/02.02.01_60/
en_3022170202v020201p.pdf.
19. Jungwoo Joh and Jesús A. Del Alamo, “A Current-Transient Method-
ology for Trap Analysis for GaN High Electron Mobility Transistors,”
IEEE Transactions on Electron Devices 58.1, 2011, pp. 132-140.
Mastering the Thermal Challenges
of Advanced Defense Subsystems
Duncan Bosworth and Gary Wenger
Analog Devices Inc., Norwood, Mass.
W
ith the continued drive to smaller
form factor, such as smaller munitions
and unmanned systems, the defense
world is pushing the boundaries of electronic
system integration and processing densities. Al-
though smaller and smaller footprints are now
becoming a reality, the challenge of thermal dis-
sipation is often not considered. Yet meeting the
thermal challenges, to ensure long term reliabil-
ity and repeatable system performance, is be-
coming a more significant
RF
part of the system design,
particularly considering the
extreme temperature rang-
es over which many aero-
space and defense systems
Digital
must operate. To meet sys-
s Fig. 1 A receiver module with integrated tem size, weight and power
RF and digital subsystems. The module is (SWaP) needs, an increasing
3.25" × 0.5" × 1.4" and dissipates 27 W in proportion of system design
the RF section and 22 W in the digital section. time needs to be allocated
to the thermal challenges.
Consider a typical RF receiver and transmit-
ter (see Figure 1), which could be the basis for
a military radio, element digitization for a radar
system or a communications link for an UAV or
advanced munition. Depending on the frequen-
WWW.MWJOURNAL.COM/ARTICLES/26428
19
cy of operation and specific application, the
system requires the integration of a number of
functions and technologies to achieve optimum
performance. The RF front-end requires power
and low noise amplifiers, most likely based on
GaAs or GaN. The mixing stages, intermediate
amplifiers and synthesizers will be developed us-
ing GaAs or SiGe, with the digitizers and FPGA
nodes on CMOS. This may result in four or five
different technologies across the signal chain,
with many more variations of process geome-
tries. Integrating these can result in the need to
dissipate 50 W or more in a few square inches
and limited thermal pathways. GaN-based pow-
er amplifiers (PA), widely used in radar and elec-
tronic warfare (EW) systems, present other chal-
lenges with their system requirements and pow-
er density. For example, the PA shown in Figure
2 integrates two GaN MMICs, each dissipating
80 W. Multiple PAs are grouped close together
to increase overall power.
To optimize SWaP and cost, a thorough un-
derstanding of the thermal design is necessary
to keep the temperatures of critical components
within their operating bounds. Each technology
and application has its own challenges from a
thermal perspective, and the drive to reduced
 Splitter
Heat Sink
RF Path
Heat Sink
Combiner
s Fig. 2
Cross-section of a power amplifier combining dual, 80 W
GaN MMICs.
s Fig. 3 Die (a) and board (b) thermal simulations.
SWaP concentrates heat densities. The dissipation needs
to be reviewed from multiple perspectives, as heat gener-
ated in the channel of a MMIC flows in a continuous chain
through numerous layers and interfaces until it ultimately
reaches the ambient environment. The entire chain must be
reviewed to optimize system thermal performance, SWaP
and cost.
Although the focus on system size reduction is certainly
making the thermal challenge more complex, some relief is
available from advanced process nodes and increasing de-
vice integration. Advanced SiGe and CMOS nodes are en-
abling significant power reduction, and increasing integrat-
ed digital signal processing is increasing integration. This
supports increased functionality, often at power parity with
previous generation architectures. The higher operating
junction temperature of GaN reduces the cooling require-
ments for these individual components. However, process
node migration is not enough to meet the thermal chal-
lenges. System miniaturization is seemingly moving faster.
20
SIMULATION IS KEY
While building and testing prototypes continues to be
critical in confirming design assumptions, the development
times and high costs preclude efficient optimization based
on hardware testing. Detailed simulation is essential and
enables the rapid evaluation of multiple system variations.
Trade-offs need to be evaluated from the entire system
perspective. Multiple model levels and tools are required,
as the geometry can scale six orders of magnitude, from
sub-micron gates to meter-long housings. Heat generating
and transfer mechanisms can include conduction, convec-
tion, radiation and EM energy. Simulation enables fast per-
formance and cost trade-offs, optimizing from the device
gate to system-level component placement, part design
and material selection, as well as fan and heat sink specifi-
cation (see Figure 3).
The maximum freedom comes from system designers
who have control over the entire system chain, from the
MMIC gate to the ambient environment. This allows a com-
prehensive approach to the thermal challenges, enabling
trade-offs that may change device location and require de-
vice modifications. Accomplishing this often requires mul-
tiple models and software packages. Specialized analysis
techniques such as computational fluid dynamics (CFD) for
convection to fluids/air and electromagnetic simulations
for RF losses can be used, with a handoff between each.
For example, a rack mounted, air cooled, high power, sol-
id-state amplifier used in radar or EW systems may need
the following:
• Finite element analysis (FEA) with micron-scale meshing,
including die level and heat spreader analysis
• Electromagnetic loss analysis to determine the power
generated in RF lines
• FEA at the chassis level
• CFD analysis for airflow and convection to the ambient
environment.
The greatest temperature deltas will typically occur at
the locations of greatest heat concentration, which are ulti-
mately near the gates. Typically, 70 percent of the tempera-
ture rise from ambient to junction is within the MMIC. With
GaN power densities in radar systems now above 6 W/mm
in some cases, simulating the trade-offs is critical.
SELECTING THE RIGHT MATERIALS
The choice and use of very high thermal conductivity ma-
terials to spread the heat are obviously critical. For exam-
ple, with high power GaN die that are used in the latest PAs
for radar, the substrate is typically SiC, and the first attach
layer is AuSn solder. Over just 5 mils of material, the heat
flux density may reduce from 13,000 W/mm2 to 24 W/mm2.
As the heat continues to flow through the system, spread-
ing will continue to reduce the flux density. The choice of
materials is limited by matching the coefficient of thermal
expansion (CTE), electrical conductivity to ground and the
cost and ability to manipulate the material.
CTE mismatches can result in cracking of substrates or
delamination of bonding layers, such as solder and epoxy.
Cold storage and operating temperatures, critical aspects
of aerospace and defense performance criteria, tend to
cause the greatest CTE stresses, as the solders and epoxy
are designed for processing at elevated temperature. Even
mild delamination can have catastrophic effects on the

(a)
(b)
s Fig. 4 Thermal performance of 60 W GaN die attached to CuW
and Al (a) vs. diamond-aluminum matrix on Cu and Al (b).
thermal performance of a die, if the separation is in an area
of high heat concentration — directly under a high power
FET, for example. Comparing the hot spot temperatures on
IR images of a design to known good images of the same
design is a useful method for identifying early delamination
when evaluating new materials.
Epoxy and sintered silver manufacturers are developing
products with lower modulus of elasticity to absorb the CTE
stresses while still retaining relatively good thermal perfor-
mance. Heat conductivity close to the die is a key materi-
al research area, with extremely high thermal conductivity
materials such as diamond being reviewed. As defense
systems continue to look for reduced SWaP and cost, per-
formance trade-off decisions for cost, weight and size are
always intertwined in system architecture and thermal
trade-offs. The use of materials such as diamond com-
posites may seem hard to justify; however, using even
small pieces of these materials as die heat spreaders in
areas of high heat concentration can substantially re-
duce device temperatures and enable cost and weight
savings in other parts of the system. Figure 4 compares
two thermal simulations: a 60 W GaN die attached to a
CuW carrier on an aluminum base vs. a diamond-aluminum
matrix material carrier on a copper insert in an aluminum
base. The latter reduced the junction temperature by some
37°C, providing improved system performance and life,
while enabling other SWaP and cost tradeoffs elsewhere in
the system.
21
(a) (b)
s Fig. 5 Examples of products using the integrated thermal de-
sign methodology: “man portable,” 5 W linear PA and up-converter
(a) and 2 kW HPA (b).
In other examples, convection cooled systems, such
as rack mounted, can be challenged with large tempera-
ture differences across the heat sink base and from the
heat sink fins to the ambient air. Heat sink and fan choices
have significant cost and performance impact and must be
specified from a system perspective. For a given heat sink
volume, better performance is driven by a higher convec-
tive heat transfer rate that requires greater back pressure,
such as from tighter channels or staggered/slotted fins
that break up boundary layers. This, in turn, requires larger
and more power-hungry fans. The fan choice also impacts
performance. Axial fans are typically the easiest to design
with, providing high volume for low pressure systems, while
centrifugal fans or blowers are able to push against higher
pressure but with lower volume.
Lastly, the choice of heat sink material can significantly
impact cost. In many cases, skived copper heat sinks seem
to provide a good balance between performance and cost.
Embedded heat pipes are also excellent low weight devic-
es for enhancing the effective thermal conductivity of heat
sink baseplates, although they do not work in all environ-
ments. High g-force environments pose a particular issue.
PUTTING IT ALL TOGETHER
Although it may seem that the thermal challenges are
ominous, with many trade-offs, using a systematic approach
can achieve solutions that balance cost, size and perfor-
mance. Advanced simulations provide the backbone for
quick decisions, enabling detailed analysis from the gate in
the die to the system, as well as the impact of the heat sinks
and heat spreaders. These advanced simulations help with
trade-offs, from materials choices to cooling techniques and
optimal layout. Applying these techniques with appropriate
design decisions can yield systems with high heat concen-
tration. Figure 5 shows two examples: a 37 W linear PA with
integrated up-converter and a 2 kW solid-state high power
amplifier (HPA). Both product designs utilized detailed sim-
ulation, which led to careful component integration, layout
and material selection to balance performance and cost.
The up-converter was designed to be “man portable,” so
minimizing size and weight was an important requirement.
Having the system and MMIC designed by the same team
helped in balancing reliability, cost and performance. ■
Your partners in performance
for mission critical RF systems

9-10.5 GHz GaN FEM for X-Band Radar Applications

This GaN FEM provides 4 functions in a single compact
package: T/R switch, PA, LNA and limiter. The Rx path
QPM2637 offers 21 dB gain with low noise figure of 2.7 dB. The Tx
path offers a small signal gain of 23 dB, it can deliver 4 W
of saturated power with a PAE of 38%. Designed for high
www.qorvo.com
temperature environments, the QPM2637 supports next-
generation AESA radar. Learn More.
For samples and
orders, contact our
worldwide distributor.
Qorvo’s GaN-on-SiC RF solutions set the standard for MTTF reliability – over
10 million hours at 200º based on more than 16,000 devices with 65 million
device hours. Qorvo’s GaN enables mission critical aerospace, defense and radar
systems requiring smaller, more efficient solutions with longer operating life.

To learn how Qorvo GaN powers the systems all around you, visit www.qorvo.com/gan
© Qorvo, Inc. | 2018. QORVO is a registered trademark of Qorvo, Inc. in the U.S. and in other countries. www.rfmw.com
Going Green: High Efficiency
GaN Amplifiers
Patrick Hindle, Microwave Journal Editor
Microwave Journal Editor
I
mproving amplifier efficiency is a key goal
in many systems as PAs typically consume
the largest amount of energy of any com-
ponent. GaN has several advantages over
other technologies such as high power density,
high power, high gain and high efficiency.
In order to achieve maximum efficiency, man-
ufacturers are using modulation techniques like
envelope tracking or outphasing plus DPD to
reduce distortion. Applying these techniques
to GaN has taken high power amplifiers to the
next level for many applications and promises to
greatly reduce power consumption. We asked
several leading GaN device manufacturers to
provide examples of their highest efficiency GaN
amplifier designs as we look to a greener future.
ANALOG DEVICES - BROADBAND AND
HIGH POWER A&D GaN
Norwood, Mass.
GaN is changing the RF and microwave land-
scape across communications systems ranging
from mobile wireless networks to aerospace and
defense, with most future radar, military commu-
nications and electronic warfare systems investi-
gating its benefits. By reducing device parasitic
elements, using shorter gate lengths and using
WWW.MWJOURNAL.COM/ARTICLES/26447
23
higher operating voltages, GaN transistors have
reached higher output power densities, wider
bandwidth and improved DC to RF efficiencies.
As the size of modern systems become increas-
ingly important, GaN, with its improved power
density, is up to 6× better than GaAs and pro-
vides a reduced system footprint with higher re-
liability.
One example of the benefit of leveraging
GaN is in modern phased array radar systems.
With thousands of active elements, GaN tech-
nology is providing comprehensive solutions for
transmit-receive modules, enabling increased
power density and integration with the PA, LNA
and T/R switch all developed in GaN. The high
breakdown voltage also potentially eliminates
the need for the limiter — traditionally used to
protect the LNA — reducing component count
and area, which is critical at higher frequencies
with minimal antenna aperture spacing.
GaN’s higher operating impedance also en-
ables optimized solutions for electronic sur-
veillance and countermeasures, with a single
broadband power amplifier now able to cover
multi-octave bandwidths. For example, ADI’s
HMC7149, based on a 0.25 micron process, is
 AMPLEON – HIGH EFFICIENCY CELLULAR BASE
45
STATION GAN
Nijmegen, The Netherlands
40 In cellular base stations, the need to have higher
+85°C
PSAT (dBm)

+25°C throughput forces the use of higher modulation coding

–40°C schemes, which leads to higher peak-to-average signals. As
35
a consequence, the RF PA’s average efficiency is reduced.
30
One candidate architecture suitable to minimize this issue
is out-phasing, where highly efficient switched-mode PAs
can be used. However, its pure version has some draw-
25 backs, namely the nulling problem and response to DPD
0 0.2 0.4 0.6 0.8 1.0
Frequency (GHz) algorithms. A solution for these issues is to split the oper-
ation into two ranges: pure out-phasing and a linear mode
s Fig. 1 HMC1099 saturated output power vs. frequency for in the highest and lowest input power levels, respectively.
various temperatures. The nulling problem is solved by this method, as the zero
output power is achieved when the input power is also zero
90 and not at the expense of subtracting two high power sig-
80
nals with opposite phase.
70
Ampleon has developed an out-phasing PA prototype
60
board that uses two IAF 0.25 mm GaN transistors. A qua-
si load insensitive (QLI) class-E topology is chosen for its
PAE (%)

50
+85°C
40
+25°C
30 –40°C
20
10
0
0 0.2 0.4 0.6 0.8 1.0
Frequency (GHz)
s Fig. 2 HMC1099 power-added efficiency vs. frequency for vari-
ous temperatures.
able to support amplification across 6 to 18 GHz with min-
imal variation of output power and gain. A single PA can
now be used where previous PA performance necessitated
band partitioning.
Advanced power combining solutions incorporated with
GaN yield even stronger benefits to users. Using propri-
etary power combining methods and advanced bias con-
trol/monitoring circuits in core power combined modules,
ADI’s SSPAs provide solutions from 100 W broadband to
8 kW at X-Band, combining up to 256 MMICs. These new
system-optimized solutions for radar and electronic warfare
have high efficiency and RF power density and provide al-
ternative solutions to traditional TWTs.
Another example is a 10 W GaN amplifier which op-
erates over an instantaneous bandwidth of 0.01 to 1.1
GHz with ± 0.5 dB of gain flatness (see Figure 1). It of-
fers high saturated output power with 69 percent typical
power-added efficiency (PAE) and 18.5 dB of small signal
gain (see Figure 2). The device is designed for industrial
and portable applications where GaN technology is need-
high efficiency against load modulation. The higher har-
monic impedances of the transistors were matched inside
the commercial RF package, SOT1135, which eases the
combiner design. Since the packaged devices are not very
sensitive to higher harmonic load impedances, the combin-
er is easier to design. A standard 30 mil Rogers substrate,
RO4350B, is used for the board design.
The static CW measurements for several values of input
power and phase angle using a dedicated dual input mea-
surement system are shown in Figure 3. Each line corre-
sponds to an input power level and each dot to an out-phas-
ing angle. It is evident that in the pure out-phasing regime,
at the back-off levels, the efficiency degrades considerably.
Additionally, it is interesting to see that it is possible to obtain
the same output power level using different values of input
power and phase angle. With this unique property, the best
efficiency performance can be selected (especially in back-
off) using the appropriate combination of input power and
phase angle using a lookup table approach. The unique am-
plitude and phase points corresponding to the highest effi-
ciency are chosen and lookup tables (phase, amplitude) are
formed. This method allows not only the extension of the dy-
namic range (nulling issue), but also efficiency optimization.
The mixed mode, out-phasing PA was tested with a single
carrier W-CDMA signal seen in Figures 4a (non-linearized)
and 4b (linearized). Before the linearization, the starting
value of ACLR is equal to -28 dBc and the EVM 9 percent.
After a single pass DPD cycle, it was possible to improve
80
Drain Efficiency (%)

ed to meet the increasing demands on battery lifetime. 70

The device utilizes internal prematching to realize a sin- 60
50
gle external matching network to provide high PAE over 40
the full bandwidth and is packaged in a compact, low cost 30
5 mm × 5 mm QFN package. 20
After its acquisition of Hittite, ADI is taking aim at many 10
high performance defense and aerospace applications. The 0
38 40 42 44 46 48 50
company offers a broad portfolio of experience in everything Output Power (dBm)
from device to subsystem products and dedicated labs, pro-
duction and testing facilities. s Fig. 3 Static CW measurements for various values of input power
and phase angle.
24
 RBW 30 kHz Load-Pull Optimum @ P2dB Compression
VBW 300 kHz (1T Pulse CW 100 µs/10%, Vds = 50 V, Idq = 375 mA
Ref 20 dBm Att 15 dB SWT 100 ns

Pout (W), Drain Eff iciency (%),

10 180
0 Peak Power (W)
–10 160
–20 140

Linear Gain (dB)

–30
–40 120
–50
–60 100
–70 Peak Eff iciency (%)
80
Center 2.125 GHz 2.55 MHz Span 25.5 MHz 60
Tx Channel Linear Gain @ Peak Eff iciency (dB)
W-CDMA 3GPP FWD 40
Bandwidth 3.84 MHz Linear Gain @ Peak Eff iciency (dB)
Power 17.00 dBm 20
Adjacent Channel Lower –28.97 dB
Bandwidth 3.84 MHz 0
Upper –29.61 dB 1700 1800 1900 2000 2100 2200 2300
Spacing 5 MHz Lower –40.51 dB
Upper –41.64 dB Frequency F0 (MHz)
Alternate Channel
Bandwidth 3.84 MHz
Spacing 10 MHz s Fig. 5 Output power, gain and efficiency vs. frequency.
(a)
COM’s Gen4 GaN technology, the MAGB-101822-120B0P
RBW 30 kHz is one in a family of products in MACOM’s MAGB series.
VBW 300 kHz
Ref 20 dBm Att 15 dB SWT 100 ns The series enables competitive products with state-of-the-
10
art performance for LTE base station applications at LD-
0 MOS-like cost structures (at volume). These products have
–10
–20 been optimized to deliver high drain efficiency and linear
–30 gain and are easy to linearize using digital predistortion
–40
–50 (DPD) according to some customers.
–60
–70 These products target all the cellular bands within the
Center 2.125 GHz 2.55 MHz Span 25.5 MHz 1.8 to 3.8 GHz range and deliver significant power efficien-
Tx Channel
W-CDMA 3GPP FWD
cy improvements, in addition to package size reduction
Bandwidth 3.84 MHz over legacy LDMOS. The products are housed in a plastic
Power 17.78 dBm
Adjacent Channel Lower –51.63 dB TO-272 package, operate over 400 MHz of bandwidth, and
Bandwidth 3.84 MHz Upper –51.48 dB
Spacing 5 MHz Lower –56.20 dB support 30, 40 and 60 W cellular infrastructure applications.
Alternate Channel Upper –56.98 dB The ability of this single device to cover the cellular bands
Bandwidth 3.84 MHz and power levels from 1.8 to 2.2 GHz would require multi-
Spacing 10 MHz
(b)
ple LDMOS-based products.
The device delivers 160 W of peak output power on
s Fig. 4 W-CDMA single carrier spectrum before linearization (a) the load-pull system (fundamental tuning only), has linear
and after linearization (b). gain of 20 dB and peak efficiency of 75 percent across the
these values to -51 dBc and 1.1 percent, respectively. The full band, similar performance to the ceramic version (see
achieved efficiency is high (>65 percent) and the PA re- Figure 5). Peak efficiency can be improved to well above
sponded to linearization algorithms, which showed the po- 80 percent when the device is presented with the proper
tential of the mixed mode, out-phasing concept for future harmonic terminations. A 2× MAGB-101822-120B0P sym-
high efficiency base stations. metric Doherty amplifier optimized for Band 1 is capable
Ampleon is the spinoff of the NXP RF power business, of delivering 55 dBm of peak power. When the Doherty
as NXP acquired Freescale and kept their RF power prod- amplifier is measured with a two carrier 20 MHz LTE signal,
ucts. Ampleon is a major player in the wireless infrastructure a total of 40 MHz carrier, and at 7.5 dB back-off, the gain is
power market, developing many unique designs to meet 15.5 dB and efficiency is 55 percent. Using a commercially
the next generation of RF power: the RF energy market for available DPD kit and without any special optimization, the
applications such as microwave ovens, lighting, automobile adjacent channel power ratio (ACPR) can be easily correct-
ignition systems and food processing. ed to less than -55 dBc. The Doherty amplifier achieves a
video bandwidth of 200 MHz.
MACOM – BREAKING THE MOLD WITH HIGH The device enables the implementation of a simple sym-
EFFICIENCY GaN ON Si metric Doherty amplifier design without compromising RF
Lowell, Mass. performance, compared to less performing and complex
MACOM is the only RF GaN manufacturer that pro- asymmetric Doherty topologies needed when using LD-
duces GaN on Si devices (the others use GaN on SiC). MOS based transistors. The symmetric Doherty amplifier,
MACOM does produce devices on both substrates, put- using the MAGB-101822-120B0P, is also easy to DPD lin-
ting significant focus on GaN on Si for low cost, high earize, which has been a challenge for users of other GaN-
volume markets such as wireless infrastructure and RF based products in the market. By overcoming all the RF and
energy. MACOM recently released a GaN wideband thermal challenges of designing GaN-based products in a
D-mode transistor optimized for 1.8 to 2.2 GHz modulat- plastic molded package, this product replaces high cost
ed signal operation in cellular base station applications ceramic air cavity packages without compromising RF and
and housed in an over molded plastic package. Using MA- thermal performance. MACOM’s improved package offer-
25
ing provides further system level cost savings to the cus-
35
tomer and eliminates another barrier to full GaN adoption.
30
The MACOM Gen4 technology is enabling wireless
25
carriers to deploy the latest LTE releases and significantly

Gain (dB)
20
reduce system operating expenses at highly competitive 2.8 GHz
15
price points, with a scalable supply chain combined with 3.0 GHz
10 3.2 GHz
experienced applications and design support team. MA-
5
COM is qualifying this process for widescale production
0
and also taking aim at the RF energy market. 2 6 10 14 18 22 26
Input Power (dBm)
QORVO – HIGH PERFORMANCE GaN FOR RADAR
Greensboro, N.C.. s Fig. 7Large signal gain vs. input power: Vd = 25 V; Idq = 200
mA; PW = 100 µs, Duty Cycle = 10%.
Airborne, land and naval radar platforms continue
to push for higher PAE for the transmit PAs. Improving it is capable of supporting a variety of operating conditions
the PAE of the system can result in simplified thermal and pulse applications due to the good thermal properties.
management; power supply requirement relaxation and The high PAE is achieved by optimizing the load-side
longer Tx path pulse width operation. To meet the need impedance termination for both gain stages, and the
for higher PAE in commercial and military radar applica- transistor layout for both electrical and thermal perfor-
tions at S-Band, Qorvo has developed a family of GaN mance, as well as compensating for the package parasit-
PAs. All Qorvo S-Band GaN PAs have > 50 percent PAE. ics. Measured load-pull data, package interface models,
The latest and highest efficiency member of the S-Band nonlinear transistor models, electromagnetic simulation
family is the QPA1000, well suited to meet the needs of and thermal modeling are all used to optimize the cir-
radar applications in the 2.8 to 3.2 GHz frequency band. cuit. Package parasitics are determined through similar
It is designed using the Qorvo QGaN25 0.25 mm GaN design testing, modeling and calculation.
on SiC production process to provide >58 percent PAE and Qorvo continues to push amplifier PAE as a critical
>50 W saturated power pulsed with an input power of 25 parameter for radar applications. Operating class choice,
dBm (see Figure 6). Measured PAE exceeds 60 percent in form factor, load-pull data, model accuracy and thermal
portions of the operating band with minimal changes to considerations all play an important role in achieving op-
the saturated output power. The part is a two stage, near timal overall amplifier performance.
class B design with >22 dB large signal gain (see Figure
7). The PA is mounted in a 7 mm × 7 mm × 0.85 mm, 48 GH EFFICIENCY GaN FOR SPACE APPLICATIONS
pin molded plastic QFN surface-mount package. The PA is Osaka, Japan
designed to operate at a quiescent bias point of Vd = 25 V Sumitomo Electric has recently qualified the SGN-
and Idq = 200 mA. For verification testing the pulse width 15H150IV GaN device at high reliability levels for space ap-
and duty cycle are 100 µs and 10 percent, respectively, but plications. This third generation GaN device is allowing an
advantage over previous technologies in gain and power
levels with excellent reliability. It has 150 W (51.5 dBm) typ-
Power Added Eff iciency (%)

70 ical output power and 117 W (50.7 dBm) minimum power

65 output at 4 dB gain compression when operated at a drain
60 voltage of 50 V. That is the highest power level of space
qualified GaN device available at this frequency, covering
55
22 dBm 25 dBm 1.55 to 1.6 GHz according to the company. High imped-
50 23 dBm 26 dBm ance levels allow for simple external matching circuits to
24 dBm
45 meet all performance goals. A high linear gain of 17.5 dB
40
2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4
Frequency 1.55 GHz
Frequency (GHz) Frequency 1.575 GHz
Frequency 1.6 GHz
s Fig. 6a PAE vs. frequency: Vd = 25 V; Idq = 200 mA; PW = 100
54 100
µs, Duty Cycle = 10%.
52 90

49 50 80
Output Power (dBm)

Pout (dBm)

48 70
PAE (%)

48
46 60
47
22 dBm 25 dBm 44 50
46 23 dBm 26 dBm 42 40
24 dBm
45 40 30
44 38 20
2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 24 28 32 36 40
Frequency (GHz) Pin (dBm)

s Fig. 6b Output power vs. frequency: Vd = 25 V; Idq = 200 mA; s Fig. 8 Ouput power and PAE vs. input power: Vd = 50 V; Idq =
PW = 100 µs, Duty Cycle = 10%. 500 mA; frequency 1.55 to 1.6 GHz.
26
minimum, 18.8 dB typical, makes driving the device easi-

95% Efficient
er. It has excellent PAE of 67 percent minimum (71 percent
ET
typical has been measured in an application engineering Advanced
breadboard). Figure 8 shows output power and PAE vs.
input power.
Modem Transceiver PA
The device has excellent performance over temperature
with the classic HEMT two-slope gain shape. The inflection
point is below -20ºC, making gain constant from -40º to
-10ºC, and a slope of -0.012 dB/ºC above that point. The
same phenomenon makes the power constant from -40º to
25ºC. The power has the same slope as the gain of -0.012
dB/ºC from 25º to 85ºC. PAE is approximately linear, with a
slope of -0.068 percent per ºC.
At 150 W output power, thermal considerations are very η
important. The thermal characteristics have been optimized
in the device and housing structure, resulting in very low
thermal resistance of 0.6 ºC/W typical and 0.7 ºC/W maxi- ETAdvanced
2-Way
mum. Each device is measured for thermal resistance indi- Doherty
vidually and that data is delivered with the device. The max- Class AB
imum junction temperature is 250ºC. This high maximum
junction temperature allows for a conservative de-rating to
160ºC operating junction temperature for safe and reliable
operation. Output Back Off (dB)
The package is the same as the previous generation of
GaAs high reliability designs, and the pre-match offers sim- s Fig. 9 ETAdvanced uses discrete supply levels to modulate the
ilar impedance levels as previous lower power devices. This drain of a PA, making millions of transitions per second to optimize
eases upgrading current designs to take advantage of the power consumption.
higher output power.
For those who are attempting even higher PAE, low Cds
and excellent performance at lower voltages make this de-
vice is a good candidate for envelope tracking schemes.
Low on resistance along with high reverse breakdown volt-
age (BVgd) also makes it attractive for switch-mode de-
signs. Nonlinear models are available as well, for more pre-
cise first-pass design and evaluation of nonlinear behavior
before building the device into a design.
Primary applications would be in navigation satellites,
GPS L1 Band or Galileo E1 Band. Sumitomo continues to s Fig. 10 Eta Devices’ demonstration of LTE and MC-GSM
focus on high performance satellite, radar and space appli- performance using ETAdvanced to modulate PAs built around the
cations for its GaN products. CGH40025F with 70% average final stage efficiency.

timize power consumption. At almost every place in the

WOLFSPEED, A CREE COMPANY
ENERGY EFFICIENT GaN FOR BASE STATIONS
Durham, N.C.
Wolfspeed is enabling designers to invent wireless sys-
tems for a responsible, energy-efficient future. Eta Devices,
an MIT spin-out based in Cambridge, Mass., is pioneering
the use of supply modulation through its technology called
ETAdvanced. ETAdvanced enables base station transmit-
ters to achieve the highest efficiency at peak power and
the highest efficiency at back-off of any known technique
according to the companies. And, unlike previous technol-
ogies using supply modulation, ETAdvanced has proven to
retain its advantages for both high bandwidth (multi-carrier
LTE) and high linearity (MC-GSM) applications. When com-
bined with Wolfspeed’s GaN devices, the result has been
reported to be the highest efficiency base station amplifi-
ers.
ETAdvanced solves the key power challenge facing the
wireless industry today. As shown in Figure 9, ETAdvanced
uses discrete supply levels to modulate the drain of a PA,
making millions of transitions per second in order to op-
27
wireless ecosystem, the power consumption and the heat
dissipation of the PA are limiting factors that increase cost
and size, limit output power and range, and overtax energy
supply resources.
In February 2016 at Mobile World Congress, Eta Devic-
es demonstrated what could be the highest efficiency GaN
high power amplifier compliant to MC-GSM/LTE specifica-
tions. The linearity specification for MC-GSM is that distor-
tion products must be lower than -60 dBc. According to Eta
Devices, Cree GaN HEMTs can be used over a very wide
range of drain bias voltages. They exploit this to achieve
not only high efficiency at maximum average power, but
also achieve high efficiency at back-off. Their demonstra-
tion uses Cree’s high performance CGH40025F with high
efficiency, high reliability and consistent performance.
Figure 10 shows measurements from Eta Devices’ MWC
2016 demonstration. Eta Devices was able to demonstrate
70 percent average final stage efficiency for a fully modulat-
ed LTE carrier and >60 percent average final stage efficien-
cy for MC-GSM transmissions. These efficiency numbers
include the loss of the supply modulator.
 Wolfspeed and Eta Devices have teamed together to
design highly efficient amplifiers for base station applica-
tions that promise to greatly reduce power consumption,
saving wireless communications providers millions in Op Ex
for a greener future.

CONCLUSION
It was not long ago that 50 percent efficient power am-
plifiers were the best in the industry, but now, many manu-
facturers are obtaining 65 to 70 percent efficiency. GaN and
advanced modulation and matching techniques are pushing
efficiencies higher every year. ■

Gallium nitride (GaN) technology has become a staple for high

power amplifiers used in communication systems and radar
applications. Visit Boonton.com/GaN to learn why
peak power meters are the ideal instruments for characterizing
your GaN amplifier and why Boonton has been a leader in power
measurement for over 70 years.

28
A Wideband High Efficiency
Doherty Power Amplifier Based on
Coupled Line Architecture
Guangping Xie , Zongxi Tang, Biao Zhang and Xin Cao
University of Electronic Science and Technology of China

A wideband high efficiency Doherty power amplifier (DPA) that works in the 1.85 to 2.4 GHz band
(26 percent fractional bandwidth) employs two pairs of anti-coupled lines with stepped impedance
resonators to replace the λ/4 transmission lines in a conventional DPA for harmonic suppression and
phase compensation. Maximum output power is between 43.0 and 44.1 dBm. At 6 dB output power
back-off (OPBO), the maximum drain efficiency (DE) is 61.3 percent, with gain higher than 12 dB. In the
saturated output power region, the efficiency is between 67 and 76.6 percent with gain higher than 8
dB. Compared with a conventional DPA, the third-order intermodulation distortion (IMD3) is reduced by
22 dB while efficiency and gain performance are greatly improved.

I
n wireless communication basestations,
Doherty power amplifiers1 are widely used
to provide high efficiency in the presence
of modulated signals with high peak to av-
erage power ratios. The λ/4 transmission line in
a conventional DPA is used mainly to perform
impedance matching and provide phase delay
compensation. The fractional bandwidth is typ-
ically narrow (usually less than 10 percent), be-
cause the λ/4 transmission line can achieve an
optimum impedance transformation and phase
compensation only at one frequency. This limits
its application in multiband, multi-standard base
stations. Moreover, the peaking amplifier in the
DPA operates class C, and as a result, has rela-
tively poor linearity.
The λ/4 transmission line in a conventional
DPA also exhibits weak harmonic suppression.
Some strategies to address this, like compos-
ite right/left-handed transmission lines (CRLH-
TL), defected ground structures (DGS), analyt-
ical model simplification and harmonic tuning,
have been successfully reported.2–5 Fang and
Quaglia2 report a DPA with 18 percent fractional
bandwidth; however, the DE and gain at 6 dB
OPBO are only 36 percent and 6 dB, respective-
ly. For comparable linearity, CRLH-TL and DGS
were adopted with the standard DPA topology
for a power-added efficiency (PAE) at 6 dB OPBO
of about 20 percent.3,4 Zhao et al.,5 report on a
similar case with measured results showing less
obvious improvement in linearity by utilizing a
simple analytical model. A 35 percent fraction-
al bandwidth is reported by Bathich et al.,6 by
exploiting wideband filters; in this case, a stan-
dard topology is also adopted, but the Doherty
behavior is not clearly demonstrated. Sarkeshi
et al.,7 employ frequency reconfigurable match-
ing networks enabling a fractional bandwidth of
about 20 percent; however, this requires an ex-
ternal control circuit.
In this work, a compact phase compensation
architecture based on two pairs of anti-coupled
lines loaded with stepped impedance resona-
tors replaces the λ/4 transmission line in a con-
ventional DPA. This structure performs imped-
ance conversion, phase compensation, phase
correction and harmonic suppression. At the
same time, it provides an easy way to tune per-
formance by adjusting the width and length of
the stepped impedance resonators.

WWW.MWJOURNAL.COM/ARTICLES/28980
29
 Z T21

CH =
jY0 sin ( βl) [ cos (βl) Z T11 + jZ Y0 sin (βl)] +
cos ( βl) [ cos ( βl) + jY0 sin ( βl) Z T22 ]
Z T21

PHASE COMPENSATION NETWORK DH =

Even/odd mode analysis was used to analyze the axial-
ly symmetric novel and compact DPA structure shown in
Figure 1. First analyzed is the two-port network in area 1.
Zoe and Zoo represent the even mode and odd mode char-
acteristic impedances of the parallel coupled line. [ZU], [ZD]
and [ZT] represent the impedance matrices of the parallel
coupled line, the low impedance line and the two-port net-
work in area 1, respectively. The impedance matrix of the
two-port network is given by8,9
[ Z T ] = [ ZD ] + [ ZU ] =
⎡ Z T11 Z T12 ⎤
⎢ ⎥ (1)
⎢⎣ Z T21 Z T22 ⎥⎦
where
Z T11 = j ( Z oe tanθe + Z oo tanθo ) /
2 − jZ 0 cot (βlom )
Z T12 = j ( Z oe tanθe − Z oo tanθo ) /
2 − jZ 0 cot (βlom )
jY0 sin ( βl) [ jZ 0 sin (βl) Z T11 + Z cos (βl)] +
cos ( βl) [ jZ 0 sin ( βl) + cos ( βl) Z T22 ]
Z T21
and then Z =Z T11Z T22 -Z T12 Z T21
The networks in areas 2 and 3 are symmetric and have
a parallel relationship. [MF] represents the ABCD matrix of
the two-port network that consists of the structures in areas
2 and 3. By analyzing the relation between voltage and cur-
rent of the two-port networks of areas 2 and 3, [MF] is given
as follows,
⎡ A
B /2 ⎤
[MF ] ⎢ 2CH HD ⎥ (4)
⎢⎣ H H ⎥⎦
AH, BH, CH and DH are defined in Equation 2.
The input microstrip line in area 5, the output microstrip
line in area 4 and the parallel network of areas 2 and 3 have a
cascade relationship. With [MG] representing the ABCD ma-
trix of the entire structure and [MK] representing the input
microstrip line in area 5, [MG] is given as

Z T21 = j ( Z oe tanθe − Z oo tanθo ) /

[MG ] = [MK ][MF ][MK ] =
⎡ A G BG ⎤
2 − jZ 0 cot (βlom ) ⎢ ⎥ (5)
⎢⎣ CG DG ⎥⎦
Z T22 = j ( Z oe tanθe + Z oo tanθo ) /
From the relationship between the ABCD matrix and the
2 − jZ 0 cot (βlom ) scattering matrix,9 the scattering parameters of the struc-
ture are
θe and θo represent the even and odd mode electrical S21 = 2/A G + BG / Z 0+
lengths, respectively. From the relation between the ABCD CGZ 0 + DG = S21 e jϕ21 (6)
matrix and the impedance matrix,9 we derive the ABCD
matrix of the two-port network. The physical length of the S11 =
microstrip curved line in area 2 is designated l and the ( A G + BG / Z0 − CGZ0 − DG ) /
ABCD matrix of the two-port network is designated [MH]. ( A G + BG / Z0 + CGZ0 + DG ) =
These are given as follows, S11 e jϕ11 (7)
AH BH
[MH ] = CH DH
(2) Z 21 = 1/ CG = Z 21 e jψ 21 (8)
φ21, φ11 and ψ21 are the phase of S21, S11 and Z21, respec-
l = win + π (Rinner + Rout )
/2
(3) tively. AG, BG, CG and DG are given in Equation 5. Using
design specifications for |S21|, |S11|, |Z21| and φ21 and Equa-
where, Symmetry Axis
Input
AH = Win
cos ( βl) [ cos ( βl) Z T11 + jZ Y0 sin ( βl)] + Area 5
jZ 0 sin ( βl) [ cos ( βl) + jY0 sin ( βl) Z T22 ] deg
lin
Z T21
Area 2 lom
BH =
[
cos ( βl) jZ 0 sin (βl) Z T11 + Z cos (βl) + ] Sco Wcc Wom
Area 1
jZ 0 sin ( βl) [ jZ 0 sin ( βl) + cos ( βl) Z T22 ] Area 3 Rinner lco
Wcon Rout

Z T21
Area 4
CH =
jY0 sin ( βl) [ cos (βl) Z T11 + jZ Y0 sin (βl)] +
Output

cos ( βl) [ cos ( βl) + jY0 sin ( βl) Z T22 ]

Z T21
s Fig. 1 Phase compensation structure for the Doherty power
amplifier.

DH = 30
jY0 sin ( βl) [ jZ 0 sin (βl) Z T11 + Z cos (βl)] +
 –80 140

Simulated Ph
–85 120
–90 100
Conventional 90° Line
–100 Proposed Structure 80
–105 60
1.8 1.9 2.0 2.1 2.2 2.3 2.4 2.5
Frequency (GHz)
(a)

| S11 | of Proposed Structure

TABLE 1
| S21 | of Proposed Structure Z21 OF THE PROPOSED STRUCTURE
| S11 | of Conventional 90° Line
Simulated Phase of S21 (°)

–65 200 | S21 | of Conventional 90° Line

f (GHz) |Z21| Phase (°)
–70 180
–75 160 2.10 62.42 91.89
0
–80 140

Response (dB)
–85 120 –20 2.15 55.07 91.38
–90 100 –40
Conventional 90° Line
–100 Proposed Structure 80 2.20 54.76 90.84
–60
–105 60
1.8 1.9 2.0 2.1 2.2 2.3 2.4 2.5
–80 2.25 53.57 90.29
Frequency (GHz)
(a) –100 2.30 55.65 89.62
2 3 4 5 6 7
Frequency (GHz)
| S11 | of Proposed Structure
(b) 2.35 61.66 88.67
| S21 | of Proposed Structure
s Fig. 2 Simulated performance
| S11 | of Conventional ofLine
90° the phase compensation structure vs. a conventional
90° line: ∠ S21| S(a), |S Conventional
21 | of 11
| and |S22 |90°
(b).Line

tions 6–8, a compact phase compensation architecture win saturation. Gain at 6 dB OPBO is higher than 12 dB in the
based 0on two pairs of anti-coupled lines with stepped im- 1.85 to 2.4 GHz band. Gain flatness is similar at 6 dB OPBO or
when saturated. Figure 5a shows the measured and simulat-
Response (dB)

pedance –20 resonators is implemented.

Figure
–40
2 shows its simulated performance. The simulat- ed output power as a function of input power at 1.85, 2.1 and
ed transmission phase of the structure is shown in Figure 2.4 GHz. Saturated output power over frequency is shown in
–60 the magnitude of the S-parameters is shown in
2a, while Figure 5b; the saturated power is between 43 and 44.1 dBm.
Figure –802b. Figure 2b shows that the conventional λ/4 trans- The DE at saturated output power and 6 dB OPBO as a
mission
–100
line provides almost 0 dB harmonic suppression, function of frequency and output power are shown in Fig-
while second2 3harmonic
4 5 suppression
6 7 is better than 40 dB ures 6a and 6b, respectively. The DE at saturation is be-
Frequency (GHz)
and(b) third harmonic suppression is about 30 dB in the new tween 67 and 76.6 percent, and the drain efficiency at 6
structure. Table 1 shows the characteristic impedance of dB OPBO is between 35 and 61.3 percent. Measured and
the structure, which is around 50 Ω. A compact structure simulated DE at saturation are comparable; however, the
implemented by capacitor-loaded coupled lines is reported drain efficiency at 6 dB OPBO shows some differences from
by Li et al.,8 but its second and third harmonic suppression the simulated results. This is due mainly to the influence of
is only about 20 dB. Zhang et al.,10 report on a structure fabrication tolerances.
consisting of a pair of anti-coupled lines short circuited by Figure 7 shows the measured IMD3 of the DPA based
a low impedance line; however, second and third harmonic on this work versus a conventional DPA, using two tones
suppression are also limited to around 20 dB. spaced 6 MHz apart, at 1.85, 1.90, 2.08 and 2.15 GHz.
The IMD3 is improved by about 22 dB at high-power com-
REALIZATION AND CHARACTERIZATION pared to the conventional DPA. In the region where output
To verify this design, a DPA operating over a 1.85 to 2.4
GHz band was realized. It is fabricated on a Taconic sub-
Input Output
strate with copper metallization (RF35 with a relative dielec- Matching Main
Matching Offset Line
PA
tric constant εr = 3.5, substrate height h = 0.508 mm, metal Network Network

50 Ω Load
50 Ω
thickness t = 0.035 mm and loss tangent tan δ = 0.0018). Signal
The main and peaking power amplifiers are implemented Source
Proposed
1:1
with Cree CGH40010F GaN HEMTs. Power Divider
Constructure

Figure 3 shows the block diagram and photo of the im-

Impedance
plemented DPA, respectively. A Wilkinson power divider Input Output Converter
90° Line Matching Peaking
splits the input power equally. The location of the phase Network PA
Matching
Network
Offset Line Network

compensation structure is indicated in Figure 3a. The dimen-

sions of the DPA are approximately 1.5 λg × 1 λg, where λg (a)

is the guide wavelength at center frequency f0 = 2.125 GHz. Anti-Coupled Lines with Stepped
Measurements were taken in 50 MHz steps with single-tone Impedance Resonators

CW excitation. The main PA operates class A/B with VDS = 28

V and VGS = ‐2.5 V (Id = 210 mA). The peaking PA operates
class C with VDS = 28 V and VGS = ‐3.7 V. Measured and sim-
ulated power gain as a function of output power are shown in
Figure 4. Differences between measurement and simulation
at low and moderate output power are attributed to simplify-
ing assumptions used in the model for the PA output load im-
pedance. Saturated gain is higher than 8 dB, as shown in both
(b)
measurement and simulation. Figure 4b shows measured and
simulated gain as a function of frequency at 6 dB OPBO and s Fig. 3 Block diagram (a) and fabricated (b) DPA.
31
 Measured @ 1.85 GHz Measured @ 1.85 GHz 100
Measured @ 2.10 GHz Measured @ 2.10 GHz
Measured @ 2.40 GHz

Drain Efficiency (%)

Measured @ 2.40 GHz 80
Simulated @ 1.85 GHz Simulated @ 1.85 GHz
Simulated @ 2.10 GHz Simulated @ 2.10 GHz
Simulated @ 2.40 GHz 60
Simulated @ 2.40 GHz
46 40 Measured, Saturated
20 44 Simulated, Saturated

Output Power (dBm)

42 Measured, 6 dB OPBO
20
15 40 Simulated, 6 dB OPBO
Power Gain (dB)

38
36 0
1.9 2.0 2.1 2.2 2.3 2.4
10 34 Frequency (GHz)
32 (a)
30
5 28

Measured Drain Efficiency (%)

26 80
24 1.85 GHz
0 15 20 25 30 35 1.90 GHz
30 35 40 45 60 1.95 GHz
Output Power (dBm) Input Power (dBm)
(a) 2.00 GHz
(a) 2.05 GHz
50 40
Measured @ 6 dB OPBO
Saturated Output (dBm)

2.10 GHz
Simulated @ 6 dB OPBO 40 2.20 GHz
Measured 20
Measured @ Saturated Output Power 2.35 GHz
Simulated
Simulated @ Saturated Output Power 2.40 GHz
30
20 0
20 25 30 35 40 45
Output Power (dBm)
Power Gain (dB)

15 (b)
10

10 0
1.9 2.0 2.1 2.2 2.3 2.4
s Fig. 6 Measured and simulated drain
Frequency (GHz) efficiency vs. frequency at 6 dB OPBO and
(b) saturation (a); measured drain efficiency vs.
5
output power vs. frequency (b).
s Fig. 5 Measured and simulated output
0 power vs. input power (a) and saturated
1.9 2.0 2.1 2.2 2.3 2.4 output power vs. frequency (b).
Frequency (GHz)
(b)

s Fig. 4 Measured and simulated power

gain vs. output power (a) and frequency, at
6 dB OPBO and saturation (b).
power is higher than 30 dBm, the best IMD3 is ‐44 dBc al bandwidths higher than 30 percent are shown, but their
at 2.08 GHz. IMD3 performance is improved across the saturated efficiencies and efficiencies at 6 dB OPBO are
board. Note the rise in IMD3 as the output power is in- lower. The DPA described by Sun and Jansen12 achieves a
creased above 38 dBm. This is because the peaking ampli- bandwidth higher than 30 percent, but its best IMD3 is only
fier is completely turned on, generating more third-order ‐35 dBc at high-power.
intermodulation.
DPA performance from this work is compared with other CONCLUSION
reported results in Table 2.4,11-13 Several DPAs with fraction- A compact phase compensation architecture based on
two pairs of anti-coupled lines loaded with stepped im-
0
pedance resonators replaces the λ/4 transmission line in a
Proposed @ 1.85 GHz Traditional @ 2.08 GHz conventional DPA. Compared to a conventional DPA, this
Proposed @ 2.15 GHz Traditional @ 2.15 GHz design demonstrates an improvement in efficiency, gain
–10 Proposed @ 2.08 GHz Traditional @ 1.85 GHz
Proposed @ 1.90 GHz Traditional @ 1.90 GHz and IMD3.■

–20 References
1. W. H. Doherty, “A New High Efficiency Power Amplifier for Modulated
IMD3 (dBc)

Waves,” Proceedings of the Institute of Radio Engineers, Vol. 24, No. 9,

–30 22 dB September 1936, pp. 1163–1182.
2. J. M. Rubio, J. Fang, V. Camarchia, R. Quaglia, M. Pirola and G. Ghione,
“3 to 3.6 GHz Wideband GaN Doherty Power Amplifier Exploiting Output
–40 Compensation Stages,” IEEE Transactions on Microwave Theory and Tech-
niques, Vol. 60, No. 8, August 2012, pp. 2543-2548.
3. Y. C. Jeong, S. G. Jeong, J. S. Lim and S. Nam, “A New Method to Sup-
–50
press Harmonics Using Bias Line Combined by Defected Ground Structure
in Power Amplifiers,” IEEE Microwave and Wireless Components Letters,
–60 Vol.13, No.12, December 2003, pp. 538-540.
15 20 25 30 35 40 45 4. S. H. Ji, S. K. Eun, C. S. Cho, J. W. Lee and J. Kim, “Linearity Improved
Output Power (dBm) Doherty Power Amplifier Using Composite Right/Left-Handed Transmis-
sion Lines,” IEEE Microwave and Wireless Components Letters, Vol. 18,
s Fig. 7 Measured IMD3 performance vs. output power, No. 8, December 2008, pp. 533–535.
comparing the coupled line design with a conventional Doherty. 5. Y. Zhao, A. G. Metzger, P. J. Zampardi, M. Iwamoto and P. M. Asbeck,
“Linearity Improvement of HBT-Based Doherty Power Amplifiers Based on
32
 TABLE 2
COMPARISON OF REPORTED WIDEBAND DPAS
12
References 4 11 13 This Work
(Version 1)
Frequency (GHz) 3 to 3.6 1.7 to 2.4 2.2 to 3.0 1.63 to 1.98 1.85 to 2.40
Psat (dBm) 44 to 43 41.5 to 34.5 41.8 to 40.2 34 to 31 44.1 to 43.0
Saturated DE (%) 65 to 56 53 to 72 68 to 52 60 to 44 (PAE) 76.65 to 67
DE @ 6 dB OPBO (%) 54 to 38 59 to 43 53 to 30 49 to 20 (PAE) 61.3 to 35
Gain (dB) 11.2 to 6 10.1 to 8.6 8.7 to 5.5 17 to 6 14.8 to 8.0
Max IMD3 (dBc) − − −35 (IMD3 Upper) − −44.0
Type GaN HEMT GaN HEMT GaN HEMT GaN HEMT GaN HEMT

a Simple Analytical Model,” IEEE Transactions on Microwave Theory and
Techniques, Vol. 54, No. 12, December 2006, pp. 4479–4488.
6. K. Bathich, A. Markos and G. Boeck, “A Wideband GaN Doherty Amplifier
With 35 Percent Fractional Bandwidth,” Proceedings of the European Mi-
crowave Conference, September 2010, pp. 1006-1009.
7. M. Sarkeshi, O. B. Leong, and A. van Roermund, “A Novel Doherty Ampli-
fier for Enhanced Load Modulation and Higher Bandwidth,” IEEE MTT-S
International Microwave Symposium Digest, June 2008, pp. 763–766.
8. R. Li, D. I. Kim and C. M. Choi, “Compact Structure With Three Attenuation
Poles for Improving Stopband Characteristics,” IEEE Microwave and Wire-
less Components Letters, Vol. 16, No. 12, December 2006, pp. 663–665.
9. D. M. Pozar, Microwave Engineering, 3rd ed., Wiley, N.Y., 2005.
10. J. Zang, L. Li, J. Gu and X. Sun,“Compact and Harmonic Suppression
Wilkinson Power Divider With Short Circuit Anti-Coupled Line” IEEE Mi-
crowave and Wireless Components Letters, Vol. 17, No. 9, September
2007, pp. 661-663.
11. L. Piazzon, R. Giofrè, P. Colantonio and F. Giannini, “A Wideband Doherty
Architecture With 36 Percent of Fractional Bandwidth,” IEEE Microwave
and Wireless Components Letters, Vol. 23, No. 11, November 2013, pp.
626-628.
12. G. Sun and R. H. Jansen, “Broadband Doherty Power Amplifier Via Real
Frequency Technique,” IEEE Transactions on Microwave Theory and Tech-
niques, Vol. 60, No. 1, January 2012, pp. 99-111.
13. S. Watanabe, Y. Takayama, R. Ishikawa and K. Honjo, “Miniature Broad-
band Doherty Power Amplifier With a Series-Connected Load,” IEEE
Transactions on Microwave Theory and Techniques, Vol. 63, No. 2, Febru-
ary 2015, pp. 572-579.

QPM1002:
8.5-10.5 GHz GaN FEM for X-Band Radar Applications
Qorvo’s GaN FEM provides 4 functions in a single, compact package: T/R switch,
PA, LNA and limiter. The receive path offers 25 dB gain with low noise figure
of 2.2 dB. The Tx path offers a small signal gain of 33 dB, it can deliver 3 W of
saturated power with a PAE of 32%. Designed for high temperature environments,
the QPM1002 supports next-generation AESA radar.
Learn More

33